gen_spherical_mosaic — Create a spherical mosaic image.
gen_spherical_mosaic creates a spherical mosaic image MosaicImage from the input images passed in Images. The pose of the images in space, which is used to compute the position of the images with respect to the surface of the sphere, can be determined with stationary_camera_self_calibration. The camera and rotation matrices computed with stationary_camera_self_calibration can be used in CameraMatrices and RotationMatrices. A spherical mosaic can only be created from images that were taken with a stationary camera (see stationary_camera_self_calibration).
The mosaic is computed in spherical coordinates (longitude and latitude). The row axis of MosaicImage corresponds to the latitude, while the column axis corresponds to the longitude. The part of the sphere that is computed by gen_spherical_mosaic is determined by LatMin, LatMax, LongMin, and LongMax. These parameters are specified in degrees and determine a rectangular part of the latitude and longitude coordinates. The latitude -90 corresponds to the north pole (i.e., the straight up viewing direction), while 90 corresponds to the south pole (i.e., the straight down viewing direction). The longitude 0 corresponds to the straight ahead viewing direction. Negative longitudes correspond to viewing directions to the left, while positive longitudes correspond to viewing directions to the right. Hence, to obtain a complete image of the sphere, LatMin = -90, LatMax = 90, LongMin = -180, and LongMax = 180 must be used. In many cases, the mosaic will not cover the entire sphere. In these cases, it is useful to select the desired part of the sphere with the above parameters. This can be done by explicitly specifying the desired rectangle. However, often it is desirable to determine the smallest rectangle that encloses all images automatically. This can be done by using LatMin < -90, LatMax > 90, LongMin < -180, and LongMax > 180. Only the parameters that lie outside the normal range of values are determined automatically.
The angle step per pixel in MosaicImage can be selected with LatLongStep, which also is an angle specified in degrees. With this, the resolution of the mosaic image can be controlled. If LatLongStep is set to 0 the angle step is calculated automatically by trying to preserve the pixel size of the original images as well as possible.
The mode in which the images are added to the mosaic is given by StackingOrder. For StackingOrder = 'voronoi', the points in the mosaic image are determined from the Voronoi cell of the respective input image. This means that the gray values are taken from the points of the input image to whose center the pixel in the mosaic image has the smallest distance on the sphere. This mode has the advantage that vignetting and uncorrected radial distortions are less noticeable in the mosaic image because they typically are symmetric with respect to the image center. Alternatively, with the choice of parameters described described in the following, a mode can be selected that has the same effect as if the images were painted successively into the mosaic image. Here, the order in which the images are added to the mosaic image is important. Therefore, an array of integer values can be passed in StackingOrder. The first index in this array will end up at the bottom of the image stack while the last one will be on top. If 'default' is given instead of an array of integers, the canonical order (images in the order used in Images) will be used. Hence, if neither 'voronoi' nor 'default' are used, StackingOrder must contain a permutation of the numbers 1,...,n, where n is the number of images passed in Images. It should be noted that the mode 'voronoi' cannot always be used. For example, at least two images must be passed to use this mode. Furthermore, for very special configurations of the positions of the image centers on the sphere, the Voronoi cells cannot be determined uniquely. With StackingOrder = 'blend', an additional mode is available, which blends the images of the mosaic smoothly. This way seams between the images become less apparent. The seam lines between the images are the same as in 'voronoi'. This mode leads to visually more appealing images, but requires significantly more resources. If the mode 'voronoi' or 'blend' cannot be used for whatever reason the mode is switched internally to 'default' automatically.
The parameter Interpolation selects the desired interpolation mode for creating the mosaic. 'bilinear' and 'bicubic' interpolation is available for all modes of StackingOrder. 'nearest_neighbor' is only available if StackingOrder is set to 'default' or 'voronoi'.
(Array of) 3x3 projective camera matrices that determine the internal camera parameters.
Array of 3x3 transformation matrices that determine rotation of the camera in the respective image.
Minimum latitude of points in the spherical mosaic image.
Default value: -90
Suggested values: -100, -90, -80, -70, -60, -50, -40, -30, -20, -10
Restriction: LatMin <= 90
Maximum latitude of points in the spherical mosaic image.
Default value: 90
Suggested values: 10, 20, 30, 40, 50, 60, 70, 80, 90, 100
Restriction: LatMax >= -90 && LatMax > LatMin
Minimum longitude of points in the spherical mosaic image.
Default value: -180
Suggested values: -200, -180, -160, -140, -120, -100, -90, -80, -70, -60, -50, -40, -30, -20, -10
Restriction: LongMin <= 180
Maximum longitude of points in the spherical mosaic image.
Default value: 180
Suggested values: 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 120, 140, 160, 180, 200
Restriction: LongMax >= -90 && LongMax > LongMin
Latitude and longitude angle step width.
Default value: 0.1
Suggested values: 0, 0.02, 0.05, 0.1, 0.2, 0.5, 1
Restriction: LatLongStep >= 0
Mode of adding the images to the mosaic image.
Default value: 'voronoi'
Suggested values: 'blend', 'voronoi', 'default'
Mode of interpolation when creating the mosaic image.
Default value: 'bilinear'
Suggested values: 'nearest_neighbor', 'bilinear', 'bicubic'
* For the input data to stationary_camera_self_calibration, please * refer to the example for stationary_camera_self_calibration. stationary_camera_self_calibration (4, 640, 480, 1, From, To, \ HomMatrices2D, Rows1, Cols1, \ Rows2, Cols2, NumMatches, \ 'gold_standard', \ ['focus','principal_point'], \ 'true', CameraMatrix, Kappa, \ RotationMatrices, X, Y, Z, Error) gen_spherical_mosaic (Images, MosaicImage, CameraMatrix, \ RotationMatrices, -100, 100, -200, 200, 0, \ 'default','bilinear') * Alternatively, if kappa should be determined, the following calls * can be made: stationary_camera_self_calibration (4, 640, 480, 1, From, To, \ HomMatrices2D, Rows1, Cols1, \ Rows2, Cols2, NumMatches, \ 'gold_standard', \ ['focus','principal_point','kappa'], \ 'true', CameraMatrix, Kappa, \ RotationMatrices, X, Y, Z, Error) cam_mat_to_cam_par (CameraMatrix, Kappa, 640, 480, CamParam) change_radial_distortion_cam_par ('fixed', CamParam, 0, CamParOut) gen_radial_distortion_map (Map, CamParam, CamParOut, 'bilinear') map_image (Images, Map, ImagesRect) gen_spherical_mosaic (ImagesRect, MosaicImage, CameraMatrix, \ RotationMatrices, -100, 100, -200, 200, 0, \ 'default','bilinear')
If the parameters are valid, the operator gen_spherical_mosaic returns the value 2 (H_MSG_TRUE). If necessary an exception is raised.
Lourdes Agapito, E. Hayman, I. Reid: “Self-Calibration of Rotating
and Zooming Cameras”; International Journal of Computer Vision;
vol. 45, no. 2; pp. 107--127; 2001.
Olivier Faugeras, Quang-Tuan Luong: “The Geometry of Multiple Images: The Laws That Govern the Formation of Multiple Images of a Scene and Some of Their Applications”; MIT Press, Cambridge, MA; 2001.