The main purpose of the module is to convert an image acquired with an input camera model by projecting it into an output camera module.
If the output camera model is the same as the input but without distortion coefficients and perfect parameters, the result will be called
"rectified". In order to setup a video rectifier, a rig.json with intrinsics calibration of the cameras is required.
The rectifier also allows for homography transformation during the projection onto the output model.
An homography can be used to simulate the location of a camera that is displaced from the original camera's location, effectively simulating
different views.
In order to initialize the rectifier, the two Calibrated Camera models are required. When warping, a user allocated output image, with
properties of the output camera and format of the input image is required. Currently any image as long as uint8 or float32 is supported.
The calculation of the distortion map will be done at initialization and recalculated when dwRectifier_setHomography() is called
Such distortion map will be used to calculate the warped image. Such map can be retrieved
by calling dwRectifier_getDistortionMap(), in the form of a dwImageCUDA with DW_IMAGE_FORMAT_RG_FLOAT32 format (XYXYXYXY...)
that correspond, for each pixel in the output image, to a location in the input image.