Face segmentation technology based on deep learning and level set method

Abstract

The invention discloses a face segmentation technology based on deep learning and a level set method, and solves the problem that a face segmentation algorithm performs unsatisfactorily with respect to a face image with too complicated background or uneven gray level in the prior art. Based on the characteristics of the diversity and complexity of human face shapes, blurred boundary contours of face images and complex backgrounds, a deep learning method is introduced into an image segmentation model, shape information of a face sample is learned by using a Boltzmann machine, and then, the shape information is introduced into an energy model expressed by a variational level set and Gaussian distribution fitting to realize fast and accurate face segmentation. The proposed face segmentation method has the characteristics of high efficiency, high accuracy and strong robustness, can reduce over-segmentation and under-segmentation phenomena, and is fast in matching and capable of real-time highly efficient segmentation.

Description

technical field [0001] The invention belongs to the technical field of image segmentation, and specifically relates to a face segmentation technology based on deep learning and a level set method that can reduce over-segmentation and under-segmentation, has fast matching speed, high accuracy, and can be segmented efficiently in real time. Background technique [0002] Face is an important biological feature of human beings, which contains rich feature information and structural information. As a key technology in face information processing, face segmentation has important application value in identity verification, content-based image retrieval, automatic monitoring, human-computer interaction and so on. Face segmentation mainly refers to determining the position and area of ​​the face in the image to facilitate the implementation of face detection and the measurement and description of important features of the face; it also lays a technical foundation for advanced underst...

AI Technical Summary

Problems solved by technology

2. Although the region-based active contour model utilizes global information, there is a problem that the sensitivity to the initial contour decreases to a certain extent.

Therefore, in order to avoid the situation that the execution effect of the face segmentation algorithm is not ideal for the face image with too complex background or uneven gray scale, it is necessary to improve the face segmentation technology of the prior art

Images