TOF camera-based hair segmentation method

Abstract

The invention relates to a TOF camera-based hair segmentation method which comprises the following steps: establishing a deep learning network based on a data set of a hair color map and a mask thereof; obtaining a depth map by using a TOF camera, and obtaining a variance map; optimizing the initial hair mask obtained by deep learning to obtain an optimized hair mask; and optimizing the optimizedhair mask again by using the variance map to obtain the final accurate hair mask. According to the method, the characteristic that a certain degree of noise is generated on hair during flight time isingeniously and reversely utilized, and higher-precision segmentation of the hair is achieved.

Description

technical field [0001] The invention relates to the field of three-dimensional images, in particular to a TOF camera-based hair segmentation method using noise in reverse. Background technique [0002] At present, the hair segmentation method is not very robust, mainly because the segmentation method based on RGB image is very strongly affected by the illumination and background, and the traditional image-based segmentation method cannot complete this task well, especially for high-definition images. Hair segmentation has long been a very difficult problem. [0003] With the advent of the era of artificial intelligence, people began to try to use deep learning methods to segment hair. But a very obvious problem is that there are very few data sets for hair. The main reason is that the hair of different people is ever-changing, and the boundary is very rough, which creates great difficulties in data labeling. Under the premise that the data set labeling is very difficult a...

AI Technical Summary

Problems solved by technology

But in our case, the edge of the noise obtained by only shooting the head is very sharp, but the high and low depth fluctuations of the hair part are very obvious

Images