Automatically recognises the outline of the visually centred subject in the input image, uses the subject as the foreground and then erases the background; returns the foreground subject image (4 channels) after erasing the background, suitable for background removal scenes such as people, animals, food and objects.