Learning OpenCV, 2nd Edition

virtual void apply(

const cv::Mat& src, // Input image

cv::Mat& dst // Result image

const cv::Rect& srcRoi = cv::Rect(0,0,-1,-1), // Region of interest to he processed

cv::Point dstOffs = cv::Point(0,0), // Offset to write to in destination image

bool isolated = false, // if true, use edge extrapolation

);

Internally, cv::FilterEngine::apply() just wraps up a call to

cv::FilterEngine::start() and then one big call to cv::FilterEngine::proceed(),

which then goes through the entire image. The only new detail is the argument dstOffs, which is the

destination offset. The destination offset is a point that indicates where in the destination image to write the

result of the processed ROI. If left at its default value of cv::Point(0,0), then the result data will be

written into the upper-left corner of the destination image. Otherwise, it will be offset appropriately. This is

particularly useful when you wish to process some ROI in the source image and write the results to the

corresponding location in the destination image.

Filter Engine Builders

OpenCV contains built in functions that will create filter engines for you. It is worth taking a moment to

appreciate what this really means. Creating a filter engine means both creating the filters that the engine

will apply (i.e., a box filter of a Gaussian filter, etc.), as well as