map_image — Apply a general transformation to an image.
map_image transforms an image Image using an arbitrary transformation Map which, for example, was previously generated using gen_image_to_world_plane_map or gen_radial_distortion_map. The multi-channel image Map must be organized as follows:
The height and the width of Map define the size of the output image ImageMapped. The number of channels in Map defines whether no interpolation or bilinear interpolation should be used. If Map only consists of one channel, no interpolation is applied during the transformation. This channel contains 'int4' (resp. 'int8' in HALCON XL if the value range of 'int4' is not sufficient) values that describe the geometric transformation: For each pixel in the output image ImageMapped the linearized coordinate of the pixel in the input image Image from which the gray value should be taken is stored.
If bilinear interpolation between the pixels in the input image should be applied, Map must consist of 5 channels. The first channel again consists of an 'int4' resp. 'int8' image and describes the geometric transformation. The channels 2-5 consist of an 'uint2' image each and contain the weights [0...1] of the four neighboring pixels that are used during bilinear interpolation. If the overall brightness of the output image ImageMapped should not differ from the overall brightness of the input image Image, the sum of the four unscaled weights must be 1 for each pixel. The weights [0...1] are scaled to the range of values of the 'uint2' image and therefore hold integer values from 0 to 65535.
Furthermore, the weights must be chosen in a way that the range of values of the output image ImageMapped is not exceeded. The geometric relation between the four channels 2-5 is illustrated in the following sketch:
The reference point of the four pixels is the upper left pixel. The linearized coordinate of the reference point is stored in the first channel.
It is also possible to use a Map that consists of a vector field containing absolute subpixel precise row and column coordinates (i.e., the field must be of the semantic type 'vector_field_absolute'). The two Map types described above can be converted into this type using convert_map_type. This type is the only type supported on compute devices!
The weights must be chosen in a way that the range of values of the output image ImageMapped is not exceeded.
For runtime reasons during the mapping process, it is not checked whether the linearized coordinates which are stored in the first channel of Map, lie inside the input image. Thus, it must be ensured by the user that this constraint is fulfilled. Otherwise, the program may crash!
map_image is parallelized automatically if and only if Map uses bilinear interpolation. map_image is executed on an OpenCL compute device only if the input map is of type 'coord_map_sub_pix' and if the input image does not exceed the maximum size of image objects of the selected device.
Image to be mapped.
Image containing the mapping data.
map_image returns 2 (H_MSG_TRUE) if all parameter values are correct. If necessary, an exception is raised.
gen_image_to_world_plane_map, gen_radial_distortion_map, convert_map_type