Skin color-based inner face image segmentation method

Abstract

The present invention relates to a skin color-based inner face image segmentation method. The method comprises the steps of converting a front-face image in the white light from the RGB color space into the YCrBr color space, simultaneously constructing a skin-color clustering-based elliptical model for filtering a candidate image, and obtaining a skin-color mask; based on the skin-color mask, obtaining the candidate area of a human face as an external rectangle, and generating a preliminary elliptical segmentation region for segmenting an original image; constructing a shrinkage space for judging a mask, subjecting the shrinkage space and a skin-color distribution diagram in an elliptical area to logical operation so as to obtain non-skin pixels remaining in four directions, and adopting the obtained pixels as an adaptive shrinkage coefficient; according to the shrinkage coefficient, updating the elliptical segmentation region; repeating the iteration process until the number of iterations reaches an upper limit or the shrinkage coefficient is within a specified error range; stopping the iteration process and outputting a target image. The method of the present invention is not dependent on any library file, and is strong in real-time property, high in recognition precision, short in shrinkage time-consuming duration, and accurate in inner-face region fitting effect.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a skin color-based inner face image segmentation method. Background technique [0002] At present, there are many skin detection devices on the market. The facial images captured by them have high definition, high capacity and bandwidth, which put a lot of pressure on system resources. At the same time, when detecting various indicators, due to background and The interference of pixel distortion on the edge of the face will greatly interfere with the detection accuracy. [0003] Some mainstream skin quality detectors, such as the VISIA system in the United States, use the method of manually selecting the detection area after taking an image to shield the invalid area and reduce the processing difficulty of the detection algorithm, but its operation is very complicated. After shooting, it is necessary to adjust the positioning of dozens of points on multiple areas (such ...

AI Technical Summary

Problems solved by technology

[0003] Some mainstream skin quality detectors, such as the VISIA system in the United States, use the method of manually selecting the detection area after taking an image to shield the invalid area and reduce the processing difficulty of the detection algorithm, but its operation is very complicated. After shooting, it is necessary to adjust the positioning of dozens of points on multiple areas (such as forehead, cheek, etc.), which greatly reduces the user experience from the perspective of the user or operator