Intelligent identification of replacement regions for mixing and replacing of persons in group portraits

摘要

A computing device identifies a target instance of a person in a target image from plural images, comprising plural instances of the person and a reference image with a reference instance of the person; aligns the target image and the reference image; determines a replacement region, including the reference and target instance of the person, having boundary passing through consistent pixels in the target and reference images; and generates an enhanced image by replacing a reference region from the reference image that corresponds with the replacement region with a target region from the target image that corresponds with the replacement region. Also disclosed is a system to mix and match faces in group digital portraits, with memory comprising: plural images including plural instances of a person, comprising a reference image with reference instance of the person. The system generates plural aesthetic ratings for the plural instances of the person; identifies a target instance of the person by determining the target instance has highest aesthetic rating; aligns target and reference images; determines a replacement region including the reference and target instances of the person, using a graph cut algorithm constrained to pass through consistent pixels while avoiding structure in the images; and generates an enhanced image. Alleviating imperfections, e.g. people blinking, in (group) pictures is an aim.