Research and comparison of facial color normalization methods relative to an etalon image

Abstract

This paper addresses the research of facial color normalization methods in images. The purpose of the research is to compare methods of normalizing facial color relative to an etalon image. The following methods were used for the research: Histogram matching, Color transfer, and Color transfer based on Face Parsing. The essence of the methods is demonstrated in detail by constructing block diagrams. The current application of normalization methods in applications and relevant literature sources is analyzed. The scientific novelty lies in the comparison of methods aimed at solving the problem of face color normalization relative to an etalon image, based on criterion descriptions. The research is based on the subject area and architecture studied and presented as a result of previous work. The application used is one of the microservices of the designed system. The results of this research can be used in practical tasks related to computer vision and, in particular, the processing of human face color in images or video frames. As a result of the research, an application was developed to normalize the color of faces relative to an etalon image using three methods, and the accuracy of the operations performed was calculated based on the corresponding criterion description.