CN101363962 A Translate to English (via PatBase internal translation) or
Select language to translate to (via Google):  


Title :
以透视投影法输出影像的方法
Output image in perspective projection method

Abstract :
本发明提供一种以圆柱形模式或透视投影模式输出影像的方法,该方法包括:(1)撷取一广角影像,是一圆形投影区域的影像,且格式可以为下列任一种:全圆形、圆角形矩形与全投影;(2)将广角影像看成是一种方位角模型.(3)利用镜头的视域角与该广角影像的一圆形投影区域的一半径,以在一具有多个方位角模式的影像中作选择;(4)详细定义一来源投射影像区域的一水平视域与一垂直视域,且该来源投射影像区域是方位角模式;(5)通过坐标转换,可以将方位角坐标模型转换成圆柱形坐标模型;(6)使用透视投影法取得圆柱形影像的局部透视投影影像;及(7)产生该输出影像。
The present invention provides a method in a cylindrical pattern or perspective projection model output image, the method comprising: (1) to capture a wide-angle image, is the image of a circular projection area, and the format can be any of the following: All circle, rounded rectangle shape and the entire projected; (2) the wide-angle image as a model of an azimuthal radius of a circular projection area (3) the use of the camera field of view angle of the wide-angle image to the. image having a plurality of azimuth mode of making a choice; (4) the detailed definition of a horizontal field of a source region of the projected image with a vertical horizon, the projected image and the source region is azimuth pattern; (5) coordinates conversion, you can convert models into cylindrical coordinates azimuth coordinate model; (6) the use of perspective projection to obtain a partial perspective projection image of the cylindrical image; and (7) generating the output image.

Claim(s) :

1. 一种以透视投影法模式输出影像的方法,其特征在于,该方法包括: (1)撷取一广角影像,是一圆形区域投影影像,且格式可以为下列任一种: 全圆形、圆角形矩形与全投影; (2)利用镜头的视域角与所述的广角影像的一圆形投影区域,以使用多个 方位角模式来描述所述的圆形区域投影影像; (3)使用一来源投射影像区域的一水平视域与一垂直视域,且该来源投射 影像区域是方位角模式; (4)转换所述的方位角模式的来源投射影像为一圆柱形模式的来源影像投 射影像; (5)以透视投影法取得所述的圆柱形模式的来源投射影像为一透视投影形 模式产生的输出部份区域影像;及 (6)产生所述的输出影像。

2.如权利要求1所述的方法,其特征在于,所述的广角影像所具有的视 域角不小于180度,镜头所产生的来源影像的水平视域角不小于180度,及 一垂直视域角不小于90度。

3.如权利要求1所述的方法,其特征在于,所述的广角影像所具有的视 域角大于或等于109.47122度,一投射的水平视域角大于或等于90度,及一 垂直视域角大于或等于90度,且当广角影像被一位于影像角落上的广角镜头 所撷取时,广角影像可涵盖一全部景像。

4.如权利要求1所述的方法,其特征在于,所述的圆形投影影像是一方 位角赤道线投影影像。

5.如权利要求1所述的方法,其特征在于,所述的圆形投影影像是一方 位角极点投影影像。

6.如权利要求1所述的方法,其特征在于,所述的圆形投影影像是一方 位角斜角投影影像。

7.如权利要求1所述的方法,其特征在于,所述的方位角模式包括直角 模式、蓝伯特等面积模式、等距离模式、立体模式、日晷模式与垂直透视其 中的任一种模型。

8.如权利要求1所述的方法,其特征在于,所述的输出影像是一透视投 影模式的影像。

9.如权利要求1所述的方法,其特征在于,所述的输出影像包括多个子 影像,且所述的子影像是圆柱形模式或透视投影模式。

10.如权利要求1所述的方法,其特征在于,所述的输出影像是一圆柱形 模式的影像。

11.如权利要求9所述的方法,其特征在于,步骤(3)之后还包括以下步骤: (131)使用一兴趣点坐标、一在所述的水平视域的中的一水平视域角与一 垂直视域角,与每一子影像的垂直视域; (132)转换所述的方位角模式的来源投射影像区域为一圆柱形模式的来源 投射影像区域; (133)推导出所述的子影像,且是圆柱形模式;及 (134)产生出所述的输出影像,输出影像包括多个子影像,且是圆柱形模 式或透视投影模式。

12.如权利要求9所述的方法,其特征在于,步骤(3)之后还包括以下步 骤: (131 ’)使用一水平视域角与一垂直视域角,以鱼眼广角的方位角影像投影 到相等于所述的水平视域角与垂直视域角的圆柱型影像; (132’)转换所述的方位角模式的来源投射影像区域为一圆柱形模式的全 来源投射影像区域;及 (133’)产生出所述的输出影像,输出影像包括多个子影像,且是圆柱形模 式或透视投影影像模式。

13.如权利要求10所述的方法,其特征在于,步骤(3)之后还包括以下步 骤: (141)使用一兴趣点坐标、一水平视域与在所述的水平视域和垂直视域中 的垂直视域角; (142)转换所述的方位角模式的来源投射影像区域为一圆柱形模式的来源 投射影像区域;及 (143)推导出所述的子影像,且是圆柱形模式的部分投影影像。

14.如权利要求10所述的方法,其特征在于,步骤(3)之后还包括以下步 骤: (141’)使用一水平视域角与一垂直视域角,来描述所述的影像的水平视域 与垂直视域; (142’)转换所述的方位角模式的来源投射影像区域为一圆柱形模式的来 源投射影像区域;及 (143’)推导出所述的子影像,且是圆柱形模式的投影影像。

15.如权利要求8所述的方法,其特征在于,步骤(5)之后还包括以下步 骤: (121)使用一兴趣点坐标与在所述的水平视域与垂直视域的水平视域角; (122)产生所述的输出投影,且是透视投影模式;及 (123)推导出所述的影像,且是一透视投影模式的局部投影影像。

16.如权利要求9所述的方法,其特征在于,步骤(4)之后还包括以下步 骤: (131")使用一兴趣点坐标、一在所述的水平视域的中的一水平视域角与 一在所述的垂直视域中的垂直视域角; (132”)转换所述的圆柱形模式的投射影像为一透视投影模式的来源投射 影像; (133")推导出所述的子影像,是透视投影模式的局部投影影像;及 (134”)产生出所述的输出影像,输出影像包括多个子影像,且是圆柱形模 式或是透视投影模式。

.

1 a perspective projection model output image, characterized in that the method comprises: (1) a wide-angle image capture is a circular area of the projected image, and the format can be any of the following: a full round, rounded rectangular shape with the entire projected; (2) the use of the sight angle of the lens with a wide angle circular projection area image according to a plurality of azimuth mode using said description circular area of the projection image; (3) the use of a source region of the projected image of a horizontal field with a vertical horizon, the projected image and the source region is azimuth mode; Source projection (4) converting the azimuth mode image of a cylindrical pattern of the source image projected image; (5) to obtain a perspective projection of the cylindrical source model of the projected image to the output image of a part of the regional perspective projection shaped pattern generation; and (6) to produce the output image described above.

2. The method of claim 1 or claim 2, wherein the angle of sight of the wide-angle image by having not less than 180 degrees, the level depending on the lens image produced by source field angle is not less than 180 degrees and a vertical field of view angle of not less than 90 degrees.

3. A method as claimed in claim 1, characterized in that the angle of sight of the wide-angle image by having greater than or equal to 109.47122 degrees, a projected horizontal field angle is greater than or equal to 90 degrees and a vertical field of view angle is greater than or equal to 90 degrees, and when a wide-angle lens wide-angle image is located on the corner of the captured image, the wide-angle image can cover a whole scene.

4. The method of claim 1 or claim 2, wherein the circular projection image is one bit angle equator projected image.

5. The method of claim 1 or claim 2, wherein the circular projection image is one of the corner pole position of the projected image.

6. The method of claim 1 or claim 2, wherein the circular projection image is the one bit of the oblique angle of the projected image.

7. The method of claim 1 or claim 2, characterized in that said pattern comprises a rectangular pattern azimuth, Lambert equal area mode, equidistant mode, stereo mode, mode sundial wherein any one model with a vertical perspective.

8. The method of claim 1 or claim 2, wherein said output image is an image of a perspective projection model.

9. The method of claim 1 or claim 2, characterized in that the output image comprises a plurality of sub-images, and said sub-image mode or perspective projection is a cylindrical mode.

10. The method as claimed in claim 1, wherein, wherein said output image is a cylindrical image pattern.

A level in the (131) using a point of interest coordinates,:

11 The method as claimed in claim 9, wherein, wherein, after step (3) further comprises the steps of a horizontal field angle and a vertical angle of sight of the field of view, and each sub-image of vertical sight; source of the projected image area (132) converting the azimuth pattern of a cylindrical projection mode sources image area; (133) to derive an image of the child, and is a cylindrical model; and (134) to generate an output image of said output image comprising a plurality of sub-images, and the cylindrical pattern or perspective projection mode.

.

12 The method as claimed in claim 9, wherein, wherein, after step (3) further comprises the steps of: (131 ') uses a horizontal field angle and a vertical field of view angle, fisheye to wide-angle image is projected onto the azimuth equal horizontal field angle of the image and the cylindrical vertical angle of sight; the source of the projected image area (132 ') converting the azimuth pattern of a cylindrical Full source-shaped pattern of the projected image area; and (133 ') to produce an output image, the output image comprises a plurality of sub-images, and the cylindrical pattern or perspective projection image mode.

.

13 The method of claim 10 or claim, characterized in that after step (3) further comprises the steps of: (141) using a point of interest coordinates, a horizontal field and in Vertical sight angle of the horizontal and vertical sight Perspective; source of the projected image area (142) converting the azimuth pattern of a cylindrical source model of the projected image area; and (143) to derive the He said sub-image, and is part of the projected image of the cylindrical pattern.

.

14 The method of claim 10 or claim, characterized in that after step (3) further comprises the steps of: (141 ') uses a horizontal field angle with a vertical horizon angle, horizontal field to describe the image and vertical sight; (142 ') to convert the source region of the projected image azimuth pattern for the source of the projected image area of a cylindrical pattern; and (143') Derivation the image of the child, and the projected image of the cylindrical pattern.

.

15 A method as claimed in claim 8, characterized in that, after step (5) further comprises the steps of: (121) using the horizontal coordinate of a point of interest in the subject to the horizontal field angle domain and vertical field of view; (122) to generate the output projection and a perspective projection model; and (123) deriving the image, and is a partial perspective projection image projection mode.

.

16 The method as claimed in claim 9, wherein, wherein, after step (4) further comprises the steps of: (131 ') using the coordinates of a point of interest, in the said one Vertical Horizon angle in the horizontal field of view in a horizontal field angle with a perpendicular to the Horizon; the projected image of the cylindrical pattern (132 ') into the perspective projection model is a source of the projected image; (133 ") to derive an image of the child, is the perspective projection model of local projection image; and (134") to generate an output image of said output image comprising a plurality of sub-images, and are cylindrical pattern or perspective Projection mode.


Description :

以透视投影法输出影像的方法 技术领域 本发明提供一种以圆柱形模型经采用透视投影法输出影像的方法,尤指 一种将鱼眼影像通过方位角模型描述经坐标转换成一圆柱形模型最后以透视 投影法对圆柱形影像做局部透视投影,取得输出影像。

背景技术 影像系统使用广角镜头这样的状况已经很久了。尤其是180度的鱼眼镜 头在这个行业已经是很知名的。鱼眼镜头产生的影像一般与半球形、部分半 球形、与视域有关。也就是说,广角镜头有较大的视角,这是相对于一般的 照相机镜头。而由广角镜头所攫取的影像,一般都有镜像桶状的扭曲。

在攫取影像时,为校正由鱼眼/广角镜头所造成的影像桶状扭曲,大部分 的浸入式观察器(immersive viewer)会在一来源影像的图像上执行几何式转换 程序。依据“透视投影修正(perspective correction)”或“扭曲修正(dewaiping)”, 转换程序中会扭曲影像,以修正透视投影影像。透视投影修正会基于影像观 察位置(image viewing position)的方向,调整影像至合适的透视投影影像。

目前,可以选择并防止一鱼眼影像的局部输出影像扭曲已有很好的方法 与装置了。但是,这些传统的方法与装置却受限于许多限制。举例说明,⑴ 半球形模式(hemisphere model)受到180度的限制。(2)当观察位置(viewing position)愈接近180度时,透视投影视图(perspective view)的扭曲也增加。(3) 产生超过180度的透视投影输出影像(perspective view)是不可能的。(4)以半球 形描述一真实的鱼眼镜头或广角镜头是不够的。(5)透视投影所产生的影像可 以得到部份局部的无桶状扭曲影像,但无法投影产生出函涵概全局的无桶状 扭曲标的影像,和他们之间的关系。

因此,找出一可以攫取并观察广角影像的方法与装置视有必要的。其条 件必须如下:(1)180度视域角的限制必须打破,甚且,视域角必须可以达到 360度。(2)广角镜头必须可以涵盖整个区域的影像。(3)必须有一方法可以运 用于各种不同投影特性的广角镜头上。(4)当扭曲部分已经校正过,一投影方 式可将一完整的广角影像投射成另外一种坐标型式的影像,并大幅减少输出 影像的扭曲量。

因此,找出一种可以选择并防止由一广角镜头所攫取的影像扭曲的方法 是非常重要的,且特别是经由鱼眼镜头所产生的影像。

发明内容 本发明的主要目的在于提供一种以圆柱形模式或透视模式输出影像的方 法,意即将一广角的视频影像转换成一扭曲减少的圆柱形模式或透视模式影 像。

一种以在圆柱形影像上使用透视投影产生输出影像的方法,包括:(1)撷 取一广角影像,是一圆形区域投影影像(circular projection image),且格式可以 为下列任一种:全圆形(foil circle)、圆角形矩形(rounded rectangle)与全投影(flill projection); (2)利用鱼眼/广角镜头视域角(degree of view)特性与该广角影像的 一圆柱形投影区域的一半径,以在一具有多个方位角模式(Azimuthal Mode)的 影像中作选择;(3)详细定义一来源投射影像区域的一水平视域(HF0VR; horizontal field of view range)与——垂直视域(VFOVR; vertical field of view range),且该来源投射影像区域是方位角模式;(4)转换该方位角模式的来源投 射影像区域为一圆柱形模式(Cylindrical Mode)的来源投射影像区域;⑶转换 该圆柱形模式的来源投射影像区域为一透视投影模式(Perspective Mode)的来 源投射影像区域;及(6)产生该输出影像。

本发明是有关于利用一种装置、算法与方法将一单一且扭曲的广角影像 转换成单一或多个子影像组合成的影像,再行输出。该子标的影像(Object Image)不是扭曲在坐标转换过程中大幅缩减,就是在透视投影模式下得到仿真 人眼透视效果所产生无扭曲的修正影像。而这些都是与人的眼睛所看到的作 比较。

附图说明 本发明的较佳实施例的目的、精神与便利性将配合附图说明,以进行更 深入的了解: 图1A~F是本发明的一种以圆柱形模式或透视模式输出影像的方法的六 个较佳实施例附图; 图1A-1是一极坐标方位的方位角投影附图; 图1A-1’是一赤道线方位的方位角投影附图; 图1A-2是一方位角至圆柱形投影的正切附图; 图1A-3是一方位角至圆柱形投影的正割附图; 图1B-1是一极坐标投影附图; 图1B-2是一赤道线投影附图; 图1B-3是一斜角投影附图; 图1B-4是图1A-2的另一观察方向附图; 图1B-5是图1A-3的另一观察方向附图; 图2是广角影像中不同的方位角模式与不同的视域角比较的一览表; 图3是圆柱形模式至方位角模式的反向图像投影附图; 图4A是一延展的来源广角影像附图; 图4B是一标的影像附图; 图4C是一具有详细说明的水平视域与垂直视域的标的影像附图; 图5A~D是包括了三个控制因素与一圆柱形影像; 图6是转换一方位角模式的视频影像至一圆柱形模式的视频影像的连续 图像附图; 图7是利用一图像投影技术推衍出相关新的圆柱形影像; 图8是可能的镜头投影型式与结果影像的一览表; 图9是方位角家族与蒂梭扭送对照图(Tissot distortion indicatrix)—览表; 图10是具有透视投影的方位角家族附图; 图11是一方位角模式的等面积影像与一方位角模式的等距离影像; 图12是多个立体投影附图; 图13是在一房间内观察整个区域并装设一摄影机于该方间的墙面附图; 图13-1是在一门口,墙上或是一个垂直的载具上装设一摄影机观察整个分 区附图; 图14是装设一台摄影机于一房间的一角落的一较高处与一角锥体视野空 间附图; 图15是可求得一最小视域角的公式与相关图像附图; 图16是一圆柱形模式的等距离投影与一蒂梭扭送对照图附图; 图17是一种概念,是由Nicolas Auguste Tissot量测并绘出图像扭曲的附 图; 图18是对应着一圆柱形坐标的平面的一经度角与一纬度角的计算附图; 及 图19是产生一可输出的透视投影影像的连续图像附图。

附图标号: 球状体101’ 平面102’ 极点投影或圆柱形投影121 赤道线投影(Equatorial projection) 122 斜角投影(Oblique projection) 123 反向图像(inverse map)300 圆柱形坐标(X,cp)的影像301 方位角坐标(x, y)的影像302 视域 611、612、613、614 输出视频影像1003 全圆形影像(fiill circle)801 圆角形矩形(rounded rectangle)802、 803 全投影(foil projection)804 光学CCD/CMOS传感器810 方位角投影区域811 球状体(spheroid)lOl' 经纬线(graticule)105 正交投影901、903 等面积投影902、1101 等距离投影903、1102 立体投影 904、1002、1200 日晷投影905、1001 垂直透视投影1004 正切圆柱表面103 正割圆柱表面104 方形经祎线105 房间1302 房间1402 角锥体视野空间(cone viewing space) 1401 公式1501 方向1504 直角投影901、1003 方位角等距离投影903、1102 蓝伯特方位角等面积投影902、1101 正切点1005 方位角立体投影904、1002、1200 投影 1204、1202 投影 506、505、502、706、504506、 507、508、505D、501 经度水平视域404 纬度垂直视域405 倾斜-摇摆-拉近拉远 (PAN-TILT-ZOOM) 108 具体实施方式 本发明是有关于利用一广角镜头产生的影像。如图1A-1、1A-1’、1A-2、 1B-4所示,该影像可以是一球状体101’,并投影至一平面102’。该平面102’ 通常是正切(tangent)于该球状体101’,但是横切(secant)的,如图1A-3与1B-5 所示。这些影像也是方位角模式(Azimuthal Mode)或天顶模式(Zenithal Mode) 的投影影像。正切点(the point of tangency)详细定义出投影的方位。以功能性 而言,该方位是投影的聚焦(focus)。平面投影(planar projection)的一般方位 (regular aspect)是一极点投影 121、一赤道线投影(Equatorial projection)122、与 一斜角投影(Oblique projection)123,如图1B-1、1B-2与1B-3。本发明的广角 影像是由方位角模式投影的,该方位角模式投影有如直角投影(orthographic projection)> 立体投影(stereographic projection)、曰晷投影(gnomonic projection)、 等距离投影(equal-distance projection)、等面积投影(equal-area projection)、与垂 直透视投影(vertical perspective projection),如图2所示。且该广角影像很容易 延展至其他所有的方位角模式。

介于方位角模式与圆柱形模式(Cylindrical Mode)的投影可以被定义出二 组图像公式(mapping equation)。第一组是前向公式(forward equation)或直接转 换(direct relation)。该前向公式或直接联系可转换极坐标(经度?I,纬度cp,半径 R)至笛卡儿坐标(Cartesian coordinate)(距原点的水平距离X,具原点的垂直距 离y),并且可提供一方便的比例因素(scale factor),而这个比例因素与地图的 比例是不同的。设定此处的公式是具有单一比例因素。第二组是一反向公式 (inverse equation),其是第一组的反向变换式。一反向图像(inverse map)300被 从一圆柱形坐标(X, cp)的影像301转换成一方位角坐标(x, y)的影像302,如图3 所示。本发明仅采用反向公式。

因为由鱼眼/广角镜头所产生的影像都会随着市视域角的增大伴随着严重 的桶状影像扭曲(barrel distortion),起因是与非广角影像相比较时,广角影像 的视域角会有些许增加。这样的扭曲在一平面影像上是从中心而发散对称的, 在一球状面上是从一轴而对称的,如图4A与5A所示。因此,找到一种可以 解决扭曲的广角影像是必须的,特别视广角镜头具有广大的应用特色。

本发明共利用三种投影,以推衍出标的影像(object image)。首先,该广角 视频影像(wide-angle video image)是一种方位角模式的视频投影。通过这三种 投影方式,具有的不同角度与光学特性的标的影像即衍生出来的,例如圆柱 形投影(Cylindrical projection)视频影像或透视投影投影(perspective projection) 视频影像。该三种投影将于以下内容中详细讨论。

模式(1):本模式是将方位角模式的视频影像转换成圆柱形模式的视频影 像。图6中的多个视域611、612、613、614利用圆柱形投影影像的特征攫取 每一个像素(pixel)(X,y),该像素(pixelXx,y)具有经度1与纬度(p的数值。其中, FR(x, y)与Flcp(x,y)是与每一像素(x, y)中的经度值1和纬度值q>有关的公式, 且如下表示之: X[x,y]-FlX(x, y).........(1-1),与 cp[x, y]=Fl(p(x, y).........(1-2)。

其中,x与y是笛卡儿坐标的平面,且是针对每一圆柱形投影影像像素, X与(P分别是经度与纬度上的弧度。

模式(2):此处的投影采取模式(1)中圆柱形视频影像为一来源影像经圆柱 形投影坐标转换后的影像,再利用局部透视投影技术(local image perspective view technique)以推衍出一输出视频影像1003,如图10所示。从公式(1-1)与 (1-2)中可知,每一点皆可以(X,cp)显示,并可由以下式子获得: ^[x, y]=¥2Kx, y).........(2-1),与 cp[x, y]=F2cp(x, y).........(2-2)。

其中,F2X(x,y)与F2(p(x, y)是由相关的每一个像素(x,y)的经度值1与纬度 值(p所推衍出来。

模式(3):与模式⑴相比较,本模式的投影是利用一反向图像投影(inverse mapping projection),如图3所示,而将影像301转换为影像302。模式(3)的 来源影像是采取来自模式(2)的透视投影模式的影像。每一个透视投影的影像 的像素是有关于方位角影像的经度值X与纬度值cp。以下是二反向公式(inverse equation)F3x(Xp,(pp)与 F3y(Xp, tpp): 从一圆柱形影像中定义一特定点cpp),于是一相关方位角影像的视频 影像位置(x,y)可以由以下方程是求得: X[Ap,(pp]= F3x(Ap,cpp).........(3-1),与 y[Xp, 9P]= F3y(Xp, (pp).........(3-2) o 其中,x与y是笛卡儿坐标的平面,且是方位角影像。公式(3-1)与(3-2) 是圆柱形视频影像至方位角视频影像的反向公式。

本发明提供一种从一选取的区域或减少扭曲的全区域中获取多个广角影 像并产生透视投影(perspective view)以为一标的影像的方法。请参考图1,是 本发明的步骤图示。本发明的方法包括以下步骤: (1)撷取一广角影像,是一方位角模式的圆形区域投影影像(circular projection image),且格式可以为下列任一种:全圆形(fbll circle)、圆角形矩形 (rounded rectangle)与全投影(foil projection),如图 8 所示; (2)利用一视域角(degree of view)与该广角影像的一圆柱形投影区域的一 半径,以在一具有多个方位角模式(Azimuthal Mode)的影像中作选择; (3)使用一来源投射影像区域的一水平视域(HFOVR; horizontal field of view range)与一垂直视域(VFOVR; vertical field of view range),且该来源投射 影像区域是方位角模式; (4)转换该方位角模式的来源投射影像区域为一圆柱形模式(Cylindrical Mode)的来源投射影像区域; (5)转换该圆柱形模式的来源投射影像区域为一透视投影模式(Perspective Mode)的来源投射影像区域;及 (6)产生该输出影像。

较佳地,请参考图1A,该输出影像是一透视投影影像,因此在上述步骤 (5)之后包含以下步骤: (121)使用一兴趣点坐标与在该水平视域与该垂直视域的水平视域角; (122)产生该输出投影,且是透视投影模式;及 (123)推导出该影像,且是一透视投影模式的局部投影影像。

较佳地,请参考图1B,当该输出影像包括多个子影像时,该子影像是圆 柱形模式或透视投影模式,因此在上述步骤(3)之后包含以下步骤: (131)使用一兴趣点(hot-spots)坐标、一在该水平视域的中的一水平视域 角(HFOV)与一垂直视域角(VF0V),与每一子影像的垂直视域(VF0VR); (132)转换该方位角模式的来源投射影像区域为一圆柱形模式(Cylindrical Mode)的来源投射影像区域; (133)推导出该子影像,且是圆柱形模式;及 (134)产生出该输出影像,输出影像包括多个子影像,且是圆柱形模式或 透视投影模式。

较佳地,请参考图1C,当该输出影像包括多个子影像时,该子影像是圆 柱形模式或透示型模式,因此在上述步骤(3)之后包含以下步骤: (131’)使用一水平视域角与一垂直视域角,以鱼眼广角的方位角影像投 影到相等于该水平视域角与该垂直视域角的圆柱型影像; (132’)转换该方位角模式的来源投射影像区域为一圆柱形模式的全来源 投射影像区域(source full proj ection region);及 (133’)产生出该输出影像,输出影像包括多个子影像,且是圆柱形影像或 是透视投影影像。

较佳地,请参考图1D,当该输出影像包括多个子影像时,该子影像是圆 柱形模式或透视型投影影像,因此在上述步骤(4)之后包含以下步骤: (131”)使用一兴趣点(hot-spots)坐标、一在该水平视域的中的一水平视域 角(HFOV)与一在该垂直视域中的垂直视域角(VF0V); (132”)转换该圆柱形模式的来源投射影像为一透视投影模式的来源投影 影像; (133”)推导出该子影像,且是透视投影模式的局部投影影像;及 (134”)产生出该输出影像,输出影像包括多个子影像,且是圆柱形模式或 透视投影影像。

较佳地,请参考图1E,当该输出影像是一圆柱形影像时,在上述步骤(4) 之后包含以下步骤: (141)使用一兴趣点座、一水平视域与在该水平视域和该垂直视域中的垂 直视域角; (142)转换该方位角模式的来源投射影像区域为一圆柱形模式的来源投射 影像区域;及 (143)推导出该子影像,且是圆柱形模式的部分投影影像。

较佳地,请参考图1F,当该输出影像是一圆柱形影像时,在上述步骤(3) 之后包含以下步骤: (141’)使用一水平视域角与一垂直视域角,以鱼眼广角的方位角来源影 像投影到相等于该水平视域角与该垂直视域角的圆柱型影像; (142’)转换该方位角模式的来源投射影像区域为一圆柱形模式的来源投 射影像区域;及 (143’)推导出该子影像,且是圆柱形模式的全投影影像。

一具有一广角镜头或鱼眼镜头(fisheye lens)的数字视频影像相机是一可 攫取真实世界的景象(scene)。具有一特别广角镜头视域角的数字化视频影像信 号会被转换。请参考图8,不同的投影方法导致不同的影像,如全圆形影像(foil circle)801、圆角形矩形(rounded rectangle)802、803 与全投影(flill projection)804, 所述的影像皆显示于一光学CCD/CMOS传感器810。该来源圆形区域影像可 以看做是一种方位角模式的方位角投影区域811。如图1A-1所示,该极坐标 方位(polar aspect)的多个方位角投影,如图1A-1’所示,赤道线方位的方位角 模式投影,与方位角投影区域411是一球状体(spherOid)10r的投影。所述的投 影点的多个方位角是正对着球状体101’的中心。该方位角可以中心参考点 (central reference point)与经度值、炜度值表示之。

目前有几种方位角的方位(azimuthal aspect),且其投影系统的定位 (placement)与球状体的轴是有关联的。这些方位是如图1A-1、1B-1的极坐标 方位(polar aspect),如图 1A-1'、1B-2 的赤道线方位(equatorial aspect),如图 1A-2、1A-3、1B-3、1B-4、与 1B-5 的斜角方位(oblique aspect)。极坐标方位 是正切于球状体101’的极点。赤道线方位则正切于球状体101’的赤道线。斜 角方位则正切于其它部位。本发明的较佳实施利则釆用赤道线方位。本发明 中的方位角投影剧有下列特性:(a)当描述到一方位角投影,且一中心纬度与 经度可以被清楚地定义时,一标准点(standard point)其作用如一投影的焦点。

(b)如图1A-1所示,一经纬线(gmtiCUle)105的九十度相交,所有的线皆汇聚在 中心点上,其中的经纬线是由地图上的纬度圈(parallel)与子午线(meridian)形成 的格子(grid)定义出的。(c)如图8所示,所有自该中心点发出的方向皆是“真 实方向(true direction)”。(d)扭曲图像则是绕着中心点呈圆形。

本发明利用了多个方位角投影的模型,以为来源输入视频影像。该多个 模型有:(a)如图9、10所示的正交投影901、903。(b)如图9、11所示的等面 积投影902、1101。(c)如图9、11所示的等距离投影903、1102。(d)如图9、 10、12所示的立体投影904、1002、1200o (e)如图9、10所示的日晷投影905、 1001。(f)如图10所示的垂直透视投影1004。若反向公式(3-1)与(3-2)可以推导 出来,这些投影皆可被归纳为方位角模式。

如图6与8所示,输入单元依据投射到一 CCD/CMOS(电荷耦合组件/互 补式金氧半导体)上的影像的真位置(true position),而定义出多个标的影像的 (true position)水平视域(HFOVR)611与垂直视域(VFVOR)612、一镜头的视域 角、与投影种类。方位角模式的输出标的影像受限于水平视域与垂直视域。

球状体表面的圆柱形投影被投影至一正切圆柱表面103,如图1A-2所示,或 是一正割圆柱表面104,如图1A-3所示。然后,该正切圆柱表面103或正割 圆柱表面104就延展如一平面。

在如图6中的水平视域(HF0VR)611与垂直视域(VFVOR)612与图7中的 一兴趣点被定义时,极可利用一图像投影技术将相关新的圆柱形影像推衍出 来。圆柱形模式由此而来,盖因投影的表面是一圆柱状,如该正切圆柱表面 103,如图1A-2所示,或是该正割圆柱表面104,如图1A-3所示。一基本坐 标系统的极坐标轴恰好是球状体101’与圆柱表面103的多个轴。子午线与并 行线就被镜射至一方形经纬线105。因此,圆柱形投影即可被方式所定义。于 是,回到笛卡儿坐标。所有的圆柱形投影在一参考点上时,皆由该基本坐标 系统所组成。本发明的定义如下所示: (X0, (p0)Cylindrical=(0, 0).........(5-1) 请参考图6,是本发明一较佳实施例附图。该第一较佳实施例包括了图1 中的部分步骤。第一较佳实施例的步骤包括:(601)自一广角镜头投影一方位 角的全/部分圆形区域;(602)自该方位角的全/部分圆形区域得出一方位角投影 影像;(603)依据水平视域613与垂直视域614在该方位角投影影像中选择一 局部方位角投影影像;及(604)转换该局部方位角投影影像成一局部圆柱形投 影影像,该局部圆柱形投影影像具有一新的水平视域611与一新的垂直视域 612。与原始方位角的全/部分圆形区域相比较,局部圆柱形投影影像在扭曲的 方位有显著的减少。使用者也可在水平视域与垂直视域中选择一组经度与纬 度,及一详细定义的兴趣点,以达到倾斜-摇摆-拉近拉远(PAN-TILT-ZOOM) 的特性。

请参考图2,一个完美的广角镜头,尤其是一个完美的鱼眼镜头,其该有 下列特色:(1)在一镜子中投影的影像是从中心作对称分布的;(2)延着从原点 至一标的点(objectpoint)的投影影像平面的径向距离与一角度成正比,该角度 是通过该投影影像平面的原点的一垂直线与从该原点至该标的点的一直线间 的夹角。这意味着该完美的广角镜头所涵盖的全部景象(scene)是平均的。换句 话说,影像中心的像素与影像边缘的像素间的距离是一样的。

不幸地,不同的反正切镜头(arctangentlens)有不同的特性。这些特性包括 了镜头视域角与投影特性(projection characteristic)。当视域角增加时,桶状扭 曲也增加。方位角投影、日晷投影与立体投影是模拟真实广角镜头在光学与 物理方面最好的模式。但每一种模式在不同范围的视域角都有其不同的限制, 如图2所示。

本发明可以无死角的看到整个视野(entire viewing)。请参考图13,当监视 一房间1302的整个区域且一台摄影机装设于该房间1302的墙上时,该摄影 机的镜头所捕捉到的影像可以提供其水平视域角不小于180度,且垂直视域 角不小于90度。如图13-1,鱼眼/广角镜头挂载在如车上,门口,墙上或是一个 垂直的载具上,镜头的视域角不小于180度时,可以完整无死角的监视到所 有镜头前方水平180度的景物。如图14所示,当一台摄影机装设于一房间 1402的一角落的一较高处时,一角锥体视野空间(cone viewing space)1401视 可以涵盖整个视野。该角锥体的整个体积必须涵盖三顶点(verteX)(l,0,0)、 (0,1,0), (0,0,1)。最小镜头的视域角可由公式1501求出。该公式1501是 2acos(l/sqrt(3)),结果是109.47度,其摄影机位置坐标是(0,0,0),镜头位置坐 标是(1/3,1/3,1/3),并朝着一方向1504,如图15所示。结论是这个具有视域角 不小于180度特征与一 CCD/CMOS传感器的镜头,而该传感器的水平视域不 小于180度、垂直视域不小于90度,且其摄影机是装设于一房间的墙壁上等 这样的条件是可以利用的;或是具有视域角不小于109.47度特征与一 CCD/CMOS传感器的镜头,而该传感器的水平视域不小于90度、垂直视域 不小于90度,且其摄影机是装设于一房间的角落等这样的条件也是可以利用 的,以提供一整个的视野。

因为等距离投影是所有图像中最简单的。经纬线是属于圆柱形投影121 的赤道线方位,如图1B-1所示。它是圆柱形投影,并具备有标准的纬度圈 (parallel)与子午线(meridian)。所有的子午线都具有一定的比例,所有的讳度圈 也都具有一定的比例。本发明采取的是圆柱形等距离投影,如图16所示。图 16中的附数个小圆圈叫作底索指示线(Tissot indicatrix)或扭曲的椭圆(ellipse of distortion),详见图17。图17是一种概念,是由Nicolas Auguste Tissot量测并 绘出图像扭曲(map distortion)。当然,圆柱形等距投影并非仅单纯的圆柱形投 影的转换,而是可以延展如其他的圆柱形投影,如莫卡托投影(Mercator)、米 勒投影(Miller)、等面积投影(equal-area)等。公式(1-1)与(1-2)可以用图18表示: 一在圆柱形标地平面上的画素(x,y),一相关的经度角度\与一相关的纬度 角度cp,可由以下导出,其中: w:圆柱形影像的宽度,以画素为单位; H:圆柱形影像的高度,以画素为单位; 圆柱形影像的起始经度角; cpl:圆柱形影像的起始纬度角; X2:圆柱形影像的结束经度角; ((>2:圆柱形影像的结束纬度角;与 水平视域(7-1) 垂直视域=cp2-9l (7-2) 且,相关的经度与纬度可以如下计算: X[x,y]=¥lX(x,y) 水平视域 xx/w (p[x,y]=Flcp(x,y)=tpl+垂直视域 xy/h 公式(7-1)与(7-2)所导出的是标的圆柱形影像中的画素。为产生与一来源 鱼眼/广角影像相符的圆柱形投影,从方位角模式转换到圆柱形模式的反向图 像公式(2-1)与(2-2)幵始是必要的要件。本发明使用方位角膜是以模拟该广角 圆形(wide-angle circular)投影、全投影(fbll projection)、或圆角形区域,皆是圆 柱形影像(\(p),如图8所示。因而得出该反向图像公式(2-1)与(2-2)。而反向 图像公式(2-1)与(2-2)是对应着该来源广角影像的位置(x,y)。本发明举出六个 实施例来描述反向图像公式。

(a)如图9与10所示的直角投影901与1003,方位角的直角投影的透视图 可从一无限远的处看到地球。该透视图给了这个立体的的图一个有错觉的假 象。其在尺寸与面积上的扭曲近乎投影的限制,以显示出更真实的影像,且 较其它投影更真实,但不包括透视图。方位角模式的直角投影被从该中间点 幵始的90度所限制,因此呈现半球形。反向图像公式(2-1)与(2-2)是利用方位 角模式的直角投影所导出,且如下所述: x[cp, A,]=cos(cp)sin(l-^0) y[cp, X,]=cos(cp 1 )sin(cp)-sin(cp 1) cos(q>)cos(人—入0) (8-1) (8-2) cp是纬度,X是经度,X0与是参考经度与参考纬度。

(b)方位角等距离投影903与1102:依据一等比例因素,一图像投影中二 点间的距离与每一个点与其它点的距离和该球状体上的相关距离是不同的。

事实上,方位角等距离投影是非透视投影。令⑷与(pi是该投影中心的经度 与炜度,因此,利用方位角等距离投影衍生的反向公式与0-2)则可衍生 出以下的公式: x[(p^]=kcos((p)sin(X-X0) (9-1) y[cp,A,]=k[cos((p 1 )sin(cp)-sin(cp 1 )cos(cp)cos(X—入0)] (9-2) (c)蓝伯特方位角等面积投影902与1101:蓝伯特方位角等面积投影的面 积与该球状体的面积成正比,且是一非透视投影,如图9的投影902所示。

因此,利用蓝伯特方位角等面积投影,反向公式(2-1)与(2-2)可以导衍出以下 方程序: (d)方位角立体投影904、1002、1200:方位角立体投影是一平面透视投 影,如图10所示的正切点1005对面的地球上的点所连接而成的影像轮廓。

其中,所有的子午线与纬度圈显示出来的皆是圆形弧线或直线。经纬线的交 叉是呈90度的。在赤道线投影方面,该纬度圈的曲线是以赤道为主轴而成对 称性分布。相对于中心纬度的纬度圈是一直线,其它的纬度圈是呈凹状,且 其凹向是与该直线纬度圈相同边的极地,如此形状线于分布在半球上,其半 k=c/sin(c) c是从中心幵始的角距离,且 cos(c)=sin((p 1 )sin((p)+cos({p 1) cos((p)cos(入-入0) x[cp,入]=kcos((p)sin(入-入0) y[(p,X,]=k[cos((p 1 )sin(cp)-sin(cp 1 )cos((p)cos(X,-X,0)] q>l是标准纬度圈,X0是中心经度,且 k=sqrt(2/[ 1 +sin(cp 1 )sin((p)+ cos(cp 1 )cos((p)cos(^-A,0)]) (10-1) (10-2) 径是从中心开始的90度计算。因此,当反向公式(2-1)、(2-2)利用了方位角立 体投影后,可衍生出下列公式: 已知一球状体的半径R: x[cp,^]:=kcos((p)sin(A--A,0) (11-1) y [cp,入]=k[cos(cp 1 )sin(cp)-sin(cp 1 )cos(cp)cos(X-入 0)] (11-2) 入0是中心经度,cpl是中心纬度,且 k=2R[ 1 +sin(cp 1 )sin(cp)+cos((p 1) cos(cp)cos(A,-X,0)] (e)曰晷投影:日晷投影是一种方位角投影,其是利用该球状物(Globe)的 中心为一透视点(perspective point)。所有的大圆(great circle)都是直线,除了方 位(aspect)之外。本投影对航海是非常有用的,因其大圆所标示出的航道的距 离是最短的。本投影却受限于其透视点,且无法投射出一条垂直于该中心点 的直线,或大于90度的直线。也就是说,赤道线方位无法投射出所述的轴极 (pole),极坐标方位无法投射出该赤道。因此,当反向公式(2-1)、(2-2)利用了 方位角日晷投影后,可衍生出下列公式: X0是中心经度,(pi是中心纬度,P是透视点的距离,其是球状体半径的 单位,与 x[cp,?i]=kcos((p)sin(A,-A-0) (12-1) y[cp,^]=k[cos((p 1 )sin(cp)-sin((p 1 )cos(cp)cos(A,-^0)], (12-2) 与 k=(P-l)/[P-sin(tp 1 )sin((p)-cos(cp 1 )cos(q>)cos(A,-^0)] 其中,c是投影中心点至点(x,y)的角距离,其关系是如下: cos(c)=sin(cp 1 )sin(cp)+cos(cp 1) cos(cp)cos(入一入0) ①垂直透视投影:请参考图12,该方位角模式的垂直透视投影从一距离 观察该球状物(globe),该距离较直角投影的距离还远。该透视投影正如从一卫 星或太空交通工具上看到地球全貌的效果一样。其正确度的范围将视与该球 状物的距离而定。从所有讨论的状况来看,其角度范围是小于90度的,如图 12的投影1202所示。投影1204所示是距离增加时,投影所显示的效果。因 此,当反向公式(2-1)、(2-2)利用了方位角模式的垂直透视投影后,可衍生出 下列公式: X0是中心经度,cpl是中心纬度,P是透视点的距离,其是球状体半径1400 单位,与 x[cp,X,]=kcos((p)sin(A,-X,0) (13-1) y[(p,X,] =k[cos(cp 1 )sin((p)-sin(cp 1 )cos(cp)cos(X,-X,0)] (13-2) 其中, cos(c) =sin((p l)sin((p)+cos((p 1) cos((p)cos(A,-X0) k=(P-l)/(P-cos(c)) P是在球状物(Globe)的背面,且在投影时会被抑制。

请参考图10,是描述利用一广角影像产生一透视视图的方法。

请参考图19,是描述产生一透视模式影像的步骤。在圆柱形转换 (cylindricaltransform)中,整个影像不可在单一的转换中,而只有部分才可以。

本发明利用三个控制参数产生标的影像,其包括:(a)摇摆角(pan angle),即一 兴趣点经度角(hot-spots point longitude angle),如图 5D 的投影 506 所示。(b) 俯仰角(tilt angle),即一兴趣点纬度角(hot-spots point longitude angle),如图 5D 的投影506所示。(c)水平视域角,如如图5D的投影505所示,且被视为一放 大因素,水平视域角是在标的影向垂直中心。

(g)最后一个投影是自一圆柱形影像创造出一透视模式影像。以下是本发 明原理的公式的揭示。其基础公式可用以计算出影像平面位置,以得出圆柱 形影像像素,并具有相对应的(x,y)坐标的位置。

以下皆是针对转换的变量、描述与数学运算: (14) X:经度(俯仰)角,且其值是?i=俯仰角* 71/180, cp:纬度(摇摆)角,且其值是cp=摇摆角* 180, W:标的影像宽度 h:标的影像高度 f:透视投影的影像的水平视域角,其值可如以下表示: f=视域角*tu/180 7t: (圆周率) 令A,B,M为三个3x3的矩阵,其中, r 1 0 0 A = 0 cos(入)sin(X) 0 -siu(X) cos(X) 一 r cos(cp) 0 -sin((p) B = 0 10 ^ sin(9) 0 cos((f>)— 且 M=AxB 比例=w/(2 xtan(f/2)) 令i=0到2 M[0][i]=M[0][i]/比例 M[l][i]=M[l][i]/比例 定义 NORM(a,b)=sqrt(axa+bx b) 以下代换式是简化转换与产生透视投影的影像的数学运算。

当y=0到h, u=-M[l][0]xyx(h/2)+M[2][0]; v=-M[l][l]xyx(h/2)+M[2][l]; z=-M[l ] [2] xyx(h/2)+M[2] [2] 当x=0到w, s = -M[0][0] x x x (w / 2) + u; t--M[0][2] x x x (w / 2) + z; /* Get Perspective projection point from Cylinder */ /* Longitude Angle */ X = atan2(s, t); /* Get Perspective projection point from Cylinder */ /* Latitude Angle */ q>= -atan2( v, NORM(s, t)); 结束; 结束。

公式(14)导出一圆柱形坐标的透视投影影像。为求得一正确的影像,从圆 柱形坐标转换至方位角位置是有必要的。利用反向公式(2-1)与(2-2),所有对 应的(x,y)坐标的位置都会从来源方位影像中攫取得。当所有反向程序(8)〜(13) 都加入到公式(14)中时,公式(14)会从圆柱形影像中产生一透视投影模式影像。

公式(14)也对等于从方位角模式的广角影像中产生的透视模式影像。而圆柱形 的部分则不会产生。该公式适用所有本发明中的方位角模式,其可选择不同 的反向公式。本发明提供一种自所有的方位角模式投影中导衍出透视投影影 像的方法。

具有多个标的影像的多个图像可以利用圆柱形或透视投影,如图4A、4B、 4C所示。图4A、5A~D是扭曲的广角来源影像。图4B、4C是标的影像,且 具有特定的经度水平视域404与纬度垂直视域405。图5A~D的四个图像包括 了三个控制因素与一圆柱形影像。也就是说,任何一个透视投影子标的影像 皆包括如图7的图像704的水平视域角,图5B的投影502的水平视域角,与 如图7的投影706的兴趣点(X,cp)。图5D中的投影504是组合四个子标的影像 成一单一的标的影像,并且输出的例子。图4A~C则是从方位角模式转换至 圆柱形模式的步骤。图5A~D则包括了一透视投影图像与三个标的影像。很 明显地,本发明具有以下的特点:(a)圆柱形模式的子标的影像的投影505D自 来源广角影像提供一完整的投影,其扭曲状况远少于图5A中的投影501的这 个来源影像。(b)透视的子标的影像,如投影506、507、508自来源影像提供 了透视投影影像的三个部分,且每个子标的影像包括三个控制因素,即经由 调整兴趣点、经度水平视域角、纬度垂直视域角而成倾斜-摇摆-拉近拉远 (PAN-TILT-ZOOM)108 的特性。

本发明是利用发明人丰富的经验,以极富创意的构思,设计出简单却能 充分解决现有技术的构造。因此,本发明确实符合具有新颖性与进步性的专 利要件。唯以上所述者,仅为本发明的较佳实施例,当不能以之限制本发明 范围。即大凡依本发明权利要求所做的均等变化及修饰,仍将不失本发明的 要义所在,也不脱离本发明的精神和范围,故都应视为本发明的进一步实施 状况。

In perspective projection image output TECHNICAL FIELD The present invention provides a cylindrical model by using the perspective projection image output method, especially a fish-eye image by the model described by azimuth coordinates into a Finally, the cylindrical model of a cylindrical perspective projection image do local perspective projection to obtain the output image.

BACKGROUND imaging system using such a situation has been a long wide-angle lens. Especially the 180-degree fisheye been in this industry is very well-known. Fisheye image produced generally hemispherical, part semi-spherical, with sight-related. That is, the wide-angle lens has a larger perspective, this is relative to general camera lens. And the wide-angle lens to grab images, generally mirrored barrel distortion.

When grab image for image barrel corrected by fisheye / wide-angle lens distortion caused most of the immersion viewer (immersive viewer) performs geometric conversion process on the image of a source image. Based on "perspective projection correction (perspective correction)" or "twisted amended (dewaiping)", the conversion process will distort the image to correct for perspective projection image. Based on the perspective projection image correction can observe the position (image viewing position) direction, adjust the image to the appropriate perspective projection image.

Currently, you can select the output image and to prevent local fisheye distorted image of an existing method and apparatus of the good. However, these conventional methods and apparatus was limited to a number of limitations. Illustration, ⑴ hemispherical mode (hemisphere model) is limited by 180 degrees. Twisted (2) When the viewing position (viewing position) more close to 180 degrees, perspective projection view (perspective view) also increases. (3) generate more than 180-degree perspective projection output image (perspective view) is not possible. (4) to describe a true hemispherical fisheye or wide-angle lens is not enough. Image (5) can be produced by perspective projection to obtain part of the partial image without barrel distortion, but can not be projected to produce a letter of no barrel distortion extends into the global target image, and the relationship between them.

So, finding a wide-angle image can grab and observe the visual method and apparatus necessary. Its conditions must be as follows: (1) the 180-degree sight angle must be broken, but was even, depending on the field angle must be able to reach 360 degrees. (2) wide-angle lens must be able to cover the entire area of the image. (3) There must be a way to run for a variety of different projection characteristics of a wide-angle lens. (4) When the twisted part has been corrected, a projection mode can be a complete wide-angle image projection coordinates into another type of image, and to significantly reduce the amount of distortion of the output image.

Thus, you can select and find a way to prevent image distortion by a wide-angle lens grab is very important, and in particular through a fisheye lens image produced.

The main object of the present invention is to provide a cylindrical pattern or perspective mode output image method, meaning a video image will soon be converted into a wide-angle distortion reduction cylindrical pattern or perspective mode image.

Method

Kind on the cylindrical perspective projection image generated using the output image, comprising: (1) a wide-angle image capture is a circular area of the projection image (circular projection image), and the format can be any of the following one of: all-round (foil circle), rounded rectangle shape (rounded rectangle) and the entire projected (flill projection); (2) the use of fisheye / wide-angle lens sight angle (degree of view) features and the wide-angle image of a cylindrical a radius-shaped projection region to the image having a plurality of azimuth mode (Azimuthal Mode) in making a choice; (3) a detailed definition of a horizontal field of a source region of the projected image (HF0VR; horizontal field of view range) and - Vertical Horizon (VFOVR; vertical field of view range), and the source region of the projected image is azimuth mode; (4) converting the source azimuth pattern projected image area is a cylindrical pattern (Cylindrical Mode) The Source projected image area; ⑶ conversion of the cylindrical source model of the projected image area of a perspective projection mode (Perspective Mode) sources of the projected image area; and (6) to generate the output image.

The present invention relates to the use of an apparatus and method to a single algorithm and distorted wide angle image is converted into more sub-images are combined into a single image, re-export. The sub-target image (Object Image) is not distorted in the coordinate conversion process substantially reduced, that is, in perspective projection mode simulation of the human eye perspective effect obtained without distortion correction generated image. And these are the people's eyes and see for comparison.

Brief Description of the object of the invention, the spirit and the convenience of the preferred embodiments will be described with the accompanying drawings, for better understanding: Figure 1A ~ F is a mode of the invention in a cylindrical or perspective mode Six preferred method embodiment of an output image drawings; Figure 1A-1 is a polar coordinates azimuthal orientation projection drawings; Figure 1A-1 'is a projection azimuthal orientation equator line drawings; Fig 1A- 2 is a drawing a tangent to the cylindrical projection azimuth; Figure 1A-3 is a cylindrical projection secant azimuth to the drawings; Figure 1B-1 is a polar coordinate projection drawings; Figure 1B-2 is a equator line projection drawings; FIG. 1B-3 is an oblique projection drawings; FIG. 1B-4 is another viewing direction in FIG 1A-2 of the drawings; FIG. 1B-5 is another viewing direction of FIG. 1A-3 is attached; Figure 2 is a wide-angle images in a different azimuth pattern comparison with a different angle of sight checklist; Figure 3 is a reverse image of the cylindrical projection mode to the azimuth pattern drawings; Figure 4A is a source of wide-angle image extended drawings; Fig. 4B is a subject of the image to the drawings; Fig. 4C is an image drawing level with a target horizon with a detailed description of the vertical field of view; Figure 5A ~ D is the inclusion of three control factors and a cylindrical image; Figure 6 is an azimuth mode convert video images to a cylindrical pattern of continuous image drawing video images; Figure 7 is an image projection technology using a new derivation of the associated cylindrical image; Figure 8 is a possible projection lens List type and result images; FIG. 9 is azimuth family and pedicle hook seized and turned over control diagram (Tissot distortion indicatrix) - Lanbiao; Figure 10 is a perspective projection azimuth family drawings; Figure 11 is an azimuth mode, etc. Area Imaging and azimuthal equidistant image of a pattern; Figure 12 is a plurality of three-dimensional projection drawings; Fig. 13 is to observe the entire area within a room and a camera mounted on the wall drawings between the parties; Figure 13 1 is mounted a camera to see the entire partitions drawings in a door, a wall or on a vertical carrier; Fig. 14 is a camera mounted in a corner of a room at a higher vision cone with a corner space FIG. 15 is obtained by a minimum angle of sight formulas and drawings related images;; Figures Figure 16 is a cylindrical projection equidistant mode with a pedicle hook seized and turned over control DRAWINGS; FIG. 17 is a concept, It is measured by the amount Nicolas Auguste Tissot and draw the image distorted figures; Fig. 18 corresponds with a longitude of a cylindrical coordinate plane angle and a latitude angle calculation drawings; and 19 is to produce an output perspective BRIEF successive images projected image. reference numeral: a spherical body 101 'plane 102' pole projection or cylindrical projection 121 projection equator line (Equatorial projection) 122 oblique projection (Oblique projection) 123 reverse image (inverse map) 300 cylindrical coordinates (X, image cp) images 301 azimuth coordinate (x, y) 302 Horizon 611,612,613,614 output video image 1003 full circle image (fiill circle) 801 rounded rectangle shape (rounded rectangle) 802, 803 full area projection projection (foil projection) 804 optical CCD / CMOS sensor 810 azimuthal projection area 811 spheroids (spheroid) lOl 'latitude and longitude lines (graticule) 105 901, 903 and other orthogonal projection 903,1102 902,1101 equidistant projection stereoscopic projection 904,1002,1200 Sundial 905,1001 vertical perspective projection 1004 projection tangent secant cylindrical surface of the cylindrical surface 103 by Yi line 105 104 square room 1302 room 1402 pyramid view volume (cone viewing space) 1501 1401 Formula 1504 at right angles to the direction of projection 901 1003 azimuthal equidistant projection 903,1102 Lambert azimuthal equal-area projection 902,1101 tangent point azimuth stereoscopic projection 1005 projection 506,505,502,706,504506 904,1002,1200 projection 1204,1202, 507, 508,505D, 501 404 Latitude Longitude horizontal field Horizon 405 vertical tilt - swing - zooms (PAN-TILT-ZOOM) 108 embodiment of the present invention is an image about the use of a wide-angle lens produced. FIG. 1A-1,1A-1 ', 1A-2, as shown in 1B-4, the image may be a spherical body 101', and projected onto a plane 102 '. The plane 102 'is generally tangent (tangent) to the bulb-shaped body 101', but the cross-section (secant) as shown in Figure 1A-3 and 1B-5. These images also azimuth mode (Azimuthal Mode) or zenith mode (Zenithal Mode) of the projected image. Tangent point (the point of tangency) detailed definition of the orientation of the projection. In terms of functionality, this orientation is the projection of focus (focus). Planar projection (planar projection) of general orientation (regular aspect) is a pole projection 121, a projection equator line (Equatorial projection) 122, with a diagonal projection (Oblique projection) 123, as shown in Figure 1B-1,1B-2 and 1B-3. Wide-angle images of the present invention is projected by the azimuth mode, the azimuth pattern projection like the Cartesian projection (orthographic projection) and gt; stereoscopic projection (stereographic projection), said sundial projection (gnomonic projection), equidistant projection (equal- distance projection), equal-area projection (equal-area projection), and the vertical perspective projection (vertical perspective projection), shown in Figure 2. And the wide-angle image can easily be extended to all other azimuth mode. between azimuth mode and the cylindrical mode (Cylindrical Mode) The projector can be defined as a second set of image formula (mapping equation). The first group is a forward formula (forward equation) or direct conversion (direct relation). The forward link can be converted directly to formula or polar coordinates (longitude? I, latitude cp, radius R) to a Cartesian coordinate (Cartesian coordinate) (the horizontal distance from the origin of the X, with the origin of the vertical distance y), and may provide a convenient scaling factor (scale factor), while the proportion of the scaling factor and the map is different. The setting here is a single scaling factor formula. The second group is a reverse equation (inverse equation), which is the inverse transform of the first group. A reverse image (inverse map) 300 is converted from a cylindrical coordinate (X, cp) at an azimuth angle of the image 301 coordinate (x, y) of the image 302, as shown in Figure 3. The present invention uses only reverse the formula.

Since the fisheye / wide-angle lens will produce images with increasing angle of sight of the city accompanied by serious barrel image distortion (barrel distortion), the cause is compared with a non-wide-angle image, the wide-angle view image Domain corner there was a little increase. Such distortion in the image on a plane of symmetry diverging from the center of the spherical surface of a shaft from a symmetrical, as shown in FIG. 4A and 5A. So, to find a way to solve a wide-angle image distortion is a must, especially depending on the wide-angle lens has a broad range of application features.

The present invention utilizes three projected total for inferring the underlying image (object image). First, the wide-angle video image (wide-angle video image) is an azimuth mode video projection. Through these three projection, the subject of the image with different angles and optical characteristics, that is derived from, for example (perspective projection) video images of the cylindrical projection (Cylindrical projection) video image projection or perspective projection. The three projection will be discussed in more detail in the following. mode (1): This mode is the mode of the video image is converted into a cylindrical azimuth mode of the video image. 611,612,613,614 Fig. 6 features multiple Horizon grab a cylindrical projection image of each pixel (pixel) (X, y), the pixel (pixelXx, y) has the longitude and latitude 1 (p's Numerical wherein, FR (x, y) and Flcp (x, y) is associated with each pixel (x, y) in the longitude 1 and latitude values q and gt; related formula, and the following represents the: X [ x, y] -FlX (x, y)... (1-1), and cp [x, y] = Fl (p (x, y)... (1-2). where, x and y are the Cartesian coordinate plane, and is a cylindrical projection image for each pixel, X and (P arc of latitude and longitude are on. mode (2): Here, take the projector mode (1) a cylindrical video images of a source image after image of the cylindrical projection coordinates after conversion, re-use of local perspective projection techniques (local image perspective view technique) to derivation of an output video image 1003, as shown in Fig. We understood and (1-2) from the equation (1-1), can each point are (X, cp) display, and may be obtained by the following equation: [x, y] = ¥ 2Kx, y)... (2-1), and cp [x, y] = F2cp (x, y)... (2-2). where, F2X (x, y) and F2 (p (x, y) is made for each pixel (x, y) values associated longitude and latitude values 1 (p the derivation out. mode (3): ⑴ compared with the pattern, this pattern is projected by means of a reverse image projection (inverse mapping projection), shown in Figure 3, and the image 301 is converted to image 302. Mode (3) of the source image is to take a perspective projection mode from the mode (2) of the image. Each pixel of a perspective projection of the image is about the value of azimuth image of longitude and latitude values X cp. The following is the second reverse formula (inverse equation) F3x (Xp, (pp) and F3y (Xp, tpp): Defines a specific point from a cylindrical image cpp), then an image of the video images associated azimuth position (x, y) can be calculated by the following equation: X [Ap, (pp] = F3x (Ap, cpp)... (3-1), and y [Xp, 9P] = F3y (Xp, (pp)... (3-2) o wherein, x and y are the Cartesian coordinates of the plane, and the azimuth image. Equation (3-1) and (3-2) is the reverse formula azimuth cylindrical video images to video images.

The present invention provides a method for obtaining a plurality of wide-angle images from the area or reduce distortions in the whole region and produce a selected perspective projection (perspective view) that an image of the subject method. Refer to Figure 1, is a step of the invention is illustrated. The method of the present invention comprises the following steps: (1) capture a wide-angle image, a projected image of a circular area azimuthal mode (circular projection image), and the format may be any of the following: full circular (fbll circle), rounded rectangle shape (rounded rectangle) and the entire projected (foil projection), shown in Figure 8; (2) the use of a visual field angle (degree of view) and a radius of a cylindrical projection of the wide-angle image of the area, in for in the image having a plurality of azimuth mode (Azimuthal Mode) selected; (3) the use of a horizontal field of a source region of the projected image (HFOVR; horizontal field of view range) and a vertical field of view (VFOVR; vertical field of view range), and the source region of the projected image is azimuth mode; (4) converting the source azimuth pattern projected image area is a cylindrical pattern (Cylindrical Mode) sources of the projected image area; (5) the conversion of the Source projected image area of the cylindrical pattern of a perspective projection mode (Perspective Mode) source projection image area; and (6) to generate the output image.

Preferably, refer to Figure 1A, the output image is a perspective projection image, so after the above step (5) comprises the following steps: (121) using an interest point coordinates and the horizontal and the vertical sight horizontal field angle field of view; (122) to generate the output projection and a perspective projection model; and (123) to derive the image, and the projected image is a partial perspective projection model.

Preferably, refer to Figure 1B, when the output image comprising a plurality of sub-images, the sub-image mode or cylindrical perspective projection mode, so in the above step (3) After the following steps: (131) Use A point of interest (hot-spots) coordinates, one in sight of a horizontal angle of the horizontal field of view (HFOV) and a vertical field of view angle (VF0V), with each sub-image vertical sight (VF0VR); (132) converts the source azimuth pattern projected image area is a cylindrical pattern (Cylindrical Mode) sources of the projected image area; (133) to derive the sub-image, and is a cylindrical model; and (134) to produce the output image, the output image comprising a plurality of sub-images, and the cylindrical pattern or perspective projection mode.

Preferably, refer to Figure 1C, when the output image comprising a plurality of sub-images, the sub-images are shown through a cylindrical pattern or schema, so the above step (3) After the following steps: (131 ') uses a horizontal field angle and a vertical angle of sight, with fish-eye wide-angle equal to the azimuth image projected onto the horizontal field angle image of the vertical cylindrical sight angle; (132 ') converts the azimuth Source projected image area mode is a cylindrical pattern of the whole source of the projected image area (source full proj ection region); and (133 ') to produce the output image, the output image comprising a plurality of sub-images, and are cylindrical or imaging perspective projection image.

Preferably, refer to Figure 1D, when the output image comprising a plurality of sub-images, the sub-image is a cylindrical pattern or perspective projection-type image, so in the above step (4) after comprising the following steps: (131 ") uses a point of interest (hot-spots) coordinates, one in sight of a horizontal angle of the horizontal field of view (HFOV) and a vertical angle of the vertical sight Sight (VF0V); (132 ") Conversion of the cylindrical source model of the projected image as a perspective projection image projection mode of origin; (133 ") to derive the sub-image, and is a partial perspective projection model projection image; and (134 ') to produce the output image, output image comprising a plurality of sub-images, and the cylindrical pattern or perspective projection image.

Preferably, refer to Figure 1E, when the output image is a cylindrical image when, in the above step (4) after the following steps: (141) use a POI seat, a horizontal field and in the Vertical Horizon angle in the horizontal and the vertical sight Perspective; (142) converts the azimuth pattern area is the source of the projected image of a cylindrical source model of the projected image area; and (143) to derive the sub-images, and It is part of the projected image of the cylindrical pattern.

Preferably, refer to Figure 1F, when the output image is a cylindrical image when, in the above step (3) After the following steps: (141 ') uses a horizontal field angle and a vertical angle of sight to the wide-angle fisheye source image onto an azimuth equal to the horizontal field angle and the vertical angle of sight cylindrical image; (142 ') into the source of the azimuth pattern projected image area is a cylindrical model Source projected image area; and (143 ') deduced the sub-image, and the whole image of the cylindrical projection mode.

Digital Video Imaging Camera

has a wide-angle lens or a fisheye lens (fisheye lens) is a real-world scene can grab (scene). Digitized video image signal has a special wide-angle lens sight angle will be converted. Please refer to FIG. 8, different projection methods result in different images, such as a full circular image (foil circle) 801, rounded rectangular shape (rounded rectangle) 802,803 and the entire projected (flill projection) 804, the images are displayed in an optical CCD / CMOS sensor 810. The source image can be seen as a circular area azimuthal projection region 811 is an azimuth pattern. As shown in FIG. 1A-1, the orientation of the polar coordinates (polar aspect) of the plurality of azimuthal projection, as shown in Figure 1A-1 ', the equator line azimuthal orientation pattern projection, and azimuth region 411 is a spherical projection body (spherOid) 10r projection. A plurality of azimuthal projection point is facing the spheroid 101 'center. The azimuth can be the central reference point (central reference point) and longitude values represent the values Wei.

There are several azimuth orientation (azimuthal aspect), and its projection axis positioning system (placement) and spheroids are related. The orientation is the orientation of the polar coordinates in FIG. 1A-1,1B-1 of the (polar aspect), as shown in Figure 1A-1 ', the equator line 1B-2 of the azimuth (equatorial aspect), as shown in Figure 1A-2,1A-3, 1B-3,1B-4, and 1B-5 bevel orientation (oblique aspect). Polar orientation is tangent to the spherical body 101 'pole. Equator line is tangent to the spheroid azimuth 101 'equator. Oblique angle orientation is tangent to the other parts. Lee preferred embodiment of the present invention will preclude the use of the equatorial line orientation. The present invention has the following characteristics azimuthal projection show: (a) when it is described to an azimuthal projection, and a central latitude and longitude can be clearly defined when a standard point (standard point) and its role as one of the focus of the projector.

(b) as shown in Figure 1A-1, a latitude and longitude lines (gmtiCUle) 105 intersect at right, all the lines are gathered in the center, which is composed of latitude and longitude lines of latitude circle on the map (parallel lattice (grid)) and meridian (meridian) is formed out of the definition. (C) as shown in Figure 8, in all directions from the center are given a "true direction (true direction)". (D) twisted around a center point of the image is rounded.

The present invention utilizes a plurality of azimuthal projection models, that the source of the input video image. The multiple model has: (a) shown in FIG. 9, 10, 901, 903 orthogonal projection. (B) shown in FIG. 9, 11 and other surface-area projection 902,1101. (C) as shown in Figure equidistant projection 9,11 903,1102. Sundial perspective projection 905 projection 904,1002,1200o (e) (d) in FIG. 9, 10, 12 shown in FIG. 9, 10 as shown, 1001. Vertical (f) shown in Figure 10 in perspective projection 1004. If you reverse the formula (3-1) and (3-2) can be derived, these projections could potentially have been summarized as azimuth mode.

6 and 8, the input unit based on projected onto a CCD / CMOS (charge-coupled component / complementary metal oxide semiconductor) true position (true position) on the image, and define a plurality of subject images (true position) horizontal field (HFOVR) 611 and vertical sight (VFVOR) 612, a lens sight angle, and projection categories. Output target image azimuth mode is limited to horizontal and vertical sight sight.

Cylindrical Projection

is projected spherical surface tangent to a cylindrical surface 103, as shown in FIG. 1A-2, or a cylindrical secant surface 104, as shown in FIG. 1A-3. Then, the cylindrical surface 103 tangent or secant cylindrical surface 104 extended like a plane.

When 6 horizontal horizon (HF0VR) 611 and vertical sight (VFVOR) 612 and a point of interest in Figure 7 is defined very image projection technology can take advantage of a new cylindrical related images inferring it. Cylindrical resulting pattern, Gein projection is a cylindrical surface, such as the tangent cylindrical surface 103, as shown in FIG. 1A-2, or the secant cylindrical surface 104, as shown in FIG. 1A-3. A substantially polar coordinate system axis is exactly spherical body 101 'and the cylindrical surface 103 of the plurality of axes. Meridian and parallel line will be mirrored to a 105 square latitude and longitude lines. Thus, the cylindrical projection mode can be defined. In that the return to Cartesian coordinates. When all of a cylindrical projection on a reference point, chosen by the base coordinate system formed. The present invention is defined as follows: (X0, (p0) Cylindrical = (0, 0)... (5-1) refer to Figure 6, is a preferred embodiment of the invention BRIEF The first preferred embodiment includes the step of FIG. 1 in the step portion of the first preferred embodiment comprises: (601) from a wide-angle lens projector full / partial area of a circular azimuth; (602) from the Full / part to draw a circular area azimuth azimuthal projection image; (603) selects a partial azimuthal projection image in the azimuth of the projected image in accordance with the horizontal and vertical Horizon 613 Horizon 614; and (604) Conversion The local azimuthal projection image into a partial cylindrical projection image, the partial cylindrical projection image with a new level of sight compared with 611 full / part-circular area a new and original vertical horizon 612. azimuth, partial cylindrical projection image in distorted azimuth significant reduction in the level of user can also sight and vertical visual field, select a point of interest group longitude and latitude, and a detailed definition in order to achieve tilt - swing - closer Pull away (PAN-TILT-ZOOM) properties.

Please refer to Figure 2, a perfect wide-angle lens, especially a perfect fisheye lens, its that have the following characteristics: (1) In a mirror image projected from the center symmetrically distributed; and (2) the extension of the The radial distance from the origin to a target point (objectpoint) of the projected image plane with an angle proportional to the angle is a vertical line through the origin of the image plane of the projection with the subject from the origin to the point of a straight line between angle. This means that the whole scene (scene) perfect wide-angle lens is covered by the average. In other words, the distance between the pixel and image pixel between the edge of the image center is the same.

Unfortunately, different arctangent lens (arctangentlens) have different characteristics. These features include a camera sight angle and projection characteristics (projection characteristic). When increasing angle of sight, barrel twist also increases. Azimuthal projection, projection and stereoscopic projection sundial is to simulate real wide-angle lens in the optical and physical aspects of the best model. But each model in different ranges of sight angle has different restrictions, shown in Figure 2.

The present invention can see the entire field of view without dead (entire viewing). Refer to Figure 13, when the entire region to monitor a room 1302 and a camera mounted on the wall of the room 1302, this camera lens captured images can provide its horizontal field angle is not less than 180 degrees, and Vertical Horizon angle is not less than 90 degrees. When shown in Figure 13-1, fisheye / wide-angle lens mount as the car door, a wall or a vertical carrier, the sight angle lens is not less than 180 degrees, it can be complete without monitoring to all shot dead The scene in front of 180 degrees horizontal. 14, when a camera mounted in a corner of the room 1402 of a high place, a corner space vision cone (cone viewing space) 1401, as the can cover the entire field of vision. The entire volume of the pyramid must cover the three vertices (verteX) (l, 0,0), (0,1,0), (0,0,1). The minimum sight angle lens determined by the formula 1501. The equation 1501 is 2acos (l / sqrt (3)), the result is 109.47 degrees, which is the camera position coordinates (0,0,0), the lens position coordinates are (1 / 3,1 / 3,1 / 3), 1504 and toward a direction, as shown in Figure 15. The conclusion is that having sight angle of not less than 180 features and a CCD / CMOS sensor lens and the horizontal field of view of the sensor is not less than 180 degrees, Vertical Horizon not less than 90 degrees, and the camera is installed in a room The walls of the finest such conditions can be utilized; or having sight angle of not less than 109.47 degrees characteristics and a CCD / CMOS sensor lens and the horizontal field of view of the sensor is not less than 90 degrees, not less than 90 vertical horizon degrees, and the camera is installed in such a condition a corner of the room, etc. are also can be used to provide an entire field of view.

Because equidistant projection is the simplest of all the images. Latitude and longitude lines is part of the cylindrical projection 121 of the equator line orientation, as shown in Figure 1B-1. It is a cylindrical projection, and have a standard latitude circle (parallel) and meridian (meridian). All of Meridian have a certain proportion of all the taboo of the ring also has a certain proportion. The present invention is cylindrical equidistant projection taken, shown in Figure 16. Figure 16 attached several small circle called the bottom rope indicator line (Tissot indicatrix) or distorted oval (ellipse of distortion), see Figure 17. FIG. 17 is a concept, is measured by the amount Nicolas Auguste Tissot and draw the image distortion (map distortion). Of course, the cylindrical equidistant projection is not just a simple cylindrical projection conversion, but can be extended as other cylindrical projection, such as mercator projection (Mercator), Miller projection (Miller), equal-area projection (equal- area) and the like. Equation (1-1) and (1-2) can be represented by Figure 18: a mark on the cylindrical ground plane pixel (x, y), an associated longitude angle \ associated with a latitude angle cp, by the following Export, which: w: width cylindrical image to pixels as a unit; H: height of the cylindrical image to pixels as a unit; starting longitude angle of the cylindrical image; cpl: Starting latitude angle cylindrical image; X2: End longitude angle of the cylindrical image; ((and gt; 2: End latitude angle cylindrical image; horizontal visual field (7-1) Vertical Horizon = cp2-9l (7-2) and related The longitude and latitude can be calculated as follows: X [x, y] = ¥ lX (x, y) horizontal field xx / w (p [x, y] = Flcp (x, y) = tpl Vertical Horizon xy / h Equation (7-1) and (7-2) are derived is the subject of a cylindrical image pixel. to produce a source fisheye / wide-angle image matching cylindrical projection, the conversion from the azimuth mode to the cylindrical mode Reverse image equations (2-1) and (2-2) are necessary elements 幵 start using the orientation of the present invention is to simulate the wide angle circular cornea (wide-angle circular) projection, the total projection (fbll projection), or round shaped region, are all cylindrical imaging (\ (p), shown in Figure 8. The inverse image thus obtained formula (2-1) and (2-2), while the reverse image Formula (2-1) and (2-2) is the corresponding position of the source of wide-angle image (x, y). The present invention includes six examples describe a reverse image formula.

Rectangular projection 901

(a) as shown in Figure 9 and 10, and 1003, right-angle perspective azimuthal projection can be seen from the Earth at infinity. This perspective gave the perspective of FIG like a false illusion. Its distortion in size and area near the projection limit, to show a more realistic image, and more real than the other projectors, but does not include perspective. Cartesian projection azimuth pattern is beginning from the mid-point of 90 degrees 幵 limited, thus rendering hemispherical. Reverse image Formula (2-1) and (2-2) are derived utilizing the azimuth rectangular projection pattern, and as described below: x [cp, A,] = cos (cp) sin (l- 0) y [cp, X,] = cos (cp 1) sin (cp) -sin (cp 1) cos (q and gt;) cos (human - into 0) (8-1) (8-2) cp is Lat, X is the longitude, X0 with reference to a reference latitude and longitude.

(b) azimuthal equidistant projection 903 and 1102: the proportion of the factors on the basis of first-class, an image projection distance between two points with the relevant point distance away from each other and the point of the spherical body is different.

In fact, non-azimuthal equidistant projection perspective projection. Order ⑷ with (pi is the degree of longitude and Wei projection center, therefore, the use of reverse azimuthal equidistant projection equation derived with 0-2) can be derived from the following equation: x [(p ] = kcos ((p) sin (X-X0) (9-1) y [cp, A,] = k [cos ((p 1) sin (cp) -sin (cp 1) cos (cp) cos (X- into 0)] (9-2) (c) Lambert azimuthal equal-area projection 902 and 1101: proportional to the area Lambert azimuthal equal-area projection area of the spherical body, and is a non-perspective projection, as projection 902 of FIG. 9.

Therefore, using Lambert azimuthal equal-area projection, reverse the equation (2-1) and (2-2) can guide derivative of the following side programs: (d) azimuthal stereographic projection 904,1002,1200: azimuthal stereographic projection is a perspective projection plane, as shown in the image contour point on earth tangent point 1005 shown across 10 connected together. where all meridians and latitude circles show up are all circular arcs or straight lines. Cross latitude and longitude lines are 90 degrees. In Projection equator line, curve of the circle of latitude is the equator of the spindle from the symmetry of the distribution. Relative to the central latitude latitude circle is a straight line, the other latitude circle is concave and its concave with the line of latitude circle the same side of the Arctic, so the shape of the line to be distributed in the hemisphere, its half k = c / sin (c) c is starting from the center of Jian angular distance, and cos (c) = sin ((p 1) sin ((p) cos ({p 1) cos ((p) cos (into - into 0) x [cp, into] = kcos ((p) sin (into - into 0) y [(p, X,] = k [cos ((p 1) sin (cp) -sin (cp 1) cos ((p) cos (X, -X, 0)] q and gt; l is a standard latitude circle, X0 is the center of the longitude, and k = sqrt (2 / [1 sin (cp 1) sin ((p) cos (cp 1) cos ((p) cos (- A, 0).]) (10-1) (10-2) calculated diameter is 90 degrees from the center, therefore, when the reverse equation (2-1), (2 -2) after use of the azimuthal stereographic projection, can be derived from the following formula: a spherical body radius R is known for: x [cp, ]: = kcos ((p) sin (A - A, 0) (11 -1) y [cp, into] = k [cos (cp 1) sin (cp) -sin (cp 1) cos (cp) cos (X- into 0)] (11-2) into the center of longitude is 0, cpl is the center of the latitude, and k = 2R [1 sin (cp 1) sin (cp) cos ((p 1) cos (cp) cos (A, -X, 0)] (e) said sundial Projection: sundial azimuthal projection is a projection, which is the use of the balls (Globe) is a perspective view of the center point (perspective point). All the great circle (great circle) is a straight line, in addition to the orientation (aspect) away. The projection for Navigation is very useful, because the great circle distance marked out of the fairway is the shortest. This projection was limited by its perspective point and can not be cast out of a straight line perpendicular to the center point, or greater than 90 degrees straight. That is, the equator can not be projected orientation of the axis of the pole (pole), polar coordinate azimuth of the equator can not be projected. Thus, when the reverse equation (2-1), (2-2) the use of the azimuth sundial After the projection, can be derived from the following formula: X0 is the center of the longitude, (pi is the latitude of the center, P is the distance between the point of perspective, which is the radius of the unit sphere, and x [cp, i?] = kcos ((p) sin (A, -A-0) (12-1) y [cp, ] = k [cos ((p 1) sin (cp) -sin ((p 1) cos (cp) cos (A, - 0)], (12-2) and k = (Pl) / [P-sin (tp 1) sin ((p) -cos (cp 1) cos (q and gt;) cos (A, - 0)] wherein, c is a projection center point to the point (x, y) of the angular distance, the relationship is as follows: cos (c) = sin (cp 1) sin (cp) cos (cp 1) cos (cp) cos (into one entry 0) ① vertical perspective projection: Refer to Figure 12, the vertical projection of the azimuth pattern perspective observation that balls (globe) from a distance which is relatively far from the right angle projection. The perspective projection picture of the effect of the Earth as seen from a satellite or space vehicles the same. The scope of its accuracy will depend on the distance and the ball was rather fixed. From the condition of all discussions of view, the angle range is less than 90 degrees, as shown in the projection 12, 1202. 1204 is shown in the projection distance increases, the effect of the projection shown. Thus, when the reverse equation (2-1), (2-2) the use of the vertical azimuth mode after perspective projection can be derived from the following formula: X0 is the center of longitude, cpl is the center of latitude, P is a perspective point distance, which is a spherical body radius 1,400 units, and x [cp, X,] = kcos ((p) sin (A, -X, 0) (13-1) y [(p, X,] = k [cos (cp 1) sin ((p) -sin (cp 1) cos (cp) cos (X, -X, 0)] (13-2) wherein, cos (c) = sin ((pl) sin ((p) cos ((p 1) cos ((p) cos (A, -X0) k = (Pl) / (P-cos (c)) P is on the back of the balls (Globe), and when the projection is inhibition.

Please refer to FIG. 10, it is described using a wide-angle image generation method of a perspective view.

Please refer to FIG. 19, is described step produces a perspective mode images. Cylindrical conversion (cylindricaltransform), the entire image can not be in a single conversion, but only part of it can be.

The present invention utilizes three control parameters to generate the target image, comprising: (a) swing angle (pan angle), that is a point of interest longitude angle (hot-spots point longitude angle), as shown in Fig. 5D projection 506. (B) the pitch angle (tilt angle), that is a point of interest latitude angle (hot-spots point longitude angle), 5D projection 506 shown in FIG. (C) horizontal field angle, as shown in Fig. 5D projection 505, and is considered a major factor in place, horizontal field angle to the vertical center of the target image.

(g) the last one is from a cylindrical projection image mode to create a perspective image. The following are the principles of the present invention discloses a formula. Its base formula used to calculate the position of the image plane, in order to obtain cylindrical imaging pixel, and have corresponding (x, y) coordinate location.

The following are for conversion of variables, description and math: (14) X: Longitude (pitch) angle, and its value is i = pitch angle * 71/180, cp:? Latitude (swinging) angle, and The value is cp = swing angle * 180, W: target image width h: height of the subject image f: perspective image projection horizon horizontal angle, its value can be expressed as follows: f = Sight angle * tu / 180 7t: (pi) Let A, B, M three 3x3 matrix, where, r 1 0 0 A = 0 cos (the) sin (X) 0 -siu (X) cos (X) a r cos (cp) 0 -sin ((p) B = 0 10 sin (9) 0 cos ((f and gt;) - and M = AxB ratio = w / (2 xtan (f / 2)) so that i = 0 to 2 M [ 0] [i] = M [0] [i] / ratio of M [l] [i] = M [l] [i] / ratio defined NORM (a, b) = sqrt (axa bx b) the following substitution formula is a simplified perspective conversion and generate math projected image.

When y = 0 to h, u = -M [l] [0] xyx (h / 2) M [2] [0]; v = -M [l] [l] xyx (h / 2) M [2] [l]; z = -M [l] [2] xyx (h / 2) M [2] [2] When x = 0 to w, s = -M [0] [0 ] xxx (w / 2) u; t - M [0] [2] xxx (w / 2) z; / * Get Perspective projection point from Cylinder * / / * Longitude Angle * / X = atan2 (s, t); / * Get Perspective projection point from Cylinder * / / * Latitude Angle * / q and gt; = -atan2 (v, NORM (s, t)); end; end.

Exporting a cylindrical perspective projection image coordinates

Equation (14). To obtain a correct image, the conversion from the cylindrical coordinate to the azimuth position is necessary. Using the inverse formula (2-1) and (2-2), all of the corresponding (x, y) coordinates of the position will have to grab images from sources orientation. When all the reverse procedure (8) to (13) are added to the formula (14), the equation (14) produces a perspective projection mode image from the cylindrical image.

Equation (14) is also equal to the perspective mode produces a wide angle image from the image azimuth mode. The cylindrical portion is not generated. This formula applies to all of the present invention azimuth pattern, which is optionally different reverse formula. The present invention provides a self-azimuth mode all the guide projection images derived a perspective projection method.

A plurality of image

having a plurality of subject images can be cylindrical or perspective projection, as shown in Figure 4A, 4B, 4C shown. FIG. 4A, 5A ~ D is distorted wide angle video source. Figure 4B, 4C is the subject of the image, and has a specific longitude and latitude level sight 404 405 Vertical Horizon. Figure four image 5A ~ D comprises three controlling factors and a cylindrical image. That is, any one of the sub-perspective projection images are subject include horizontal field angle of the image 704 of FIG. 7, the projection horizontal field angle 502 in Figure 5B, the projection of a point of interest as shown in Figure 7 706 (X, cp). Figure 5D projection 504 is a combination of four sub-target image into a single image of the subject, and output examples. Figure 4A ~ C is a step to a transition from the cylindrical pattern azimuth mode. Figure 5A ~ D includes a perspective projection image and the three underlying image. Obviously, the present invention has the following characteristics: a projection 505D (a) the subject of a cylindrical pattern of sub-image source from providing a complete wide-angle image projection, which distort the situation is far less than the source image in Fig. 5A projection 501. (B) sub-perspective image of the subject, such as projection 506,507,508 from the source image provides a perspective projection image of three parts, and each sub-target image includes three control factors, namely by adjusting the POI, longitude horizontal field angle, the latitude angle from vertical horizon tilt - swing - zooms (PAN-TILT-ZOOM) 108 properties.

The inventors of the present invention is the use of rich experience, creative ideas, design a simple structure able to fully solve the prior art. Accordingly, the present invention is indeed consistent with the patent requirements of novelty and progress in nature. Only those described above, only the preferred embodiment of the present invention, when it is not to limit the scope of the present invention. Generally, under this invention that is claimed in claim made equal changes and modifications will lose the essence of the invention resides not departing from the spirit and scope of the present invention, it should be regarded as a further embodiment of the present invention the situation.