The methods provided makes a full copy of the image and modifies the pixels in the copy, which for large images can be extremely memory consuming. There should be non-const versions of these methods that modify the current image without requiring twice the memory really needed (the original image is often not needed).