Image matting method and system based on depth information

Abstract

The invention discloses an image matting method and system based on depth information. The method comprises the following steps: S1, collecting a depth image and a color image containing a target; s2,aligning the depth image with the color image to obtain an aligned depth image; s3, performing threshold binarization processing and corrosion expansion processing on the aligned depth images to obtain a ternary graph; s4, processing the ternary image and the color image through an image matting algorithm to obtain an Alpha image; and S5, separating a target by using the Alpha image and the colorimage to obtain a target image. According to the invention, a user can achieve a matting function of a real-time video without replacing a 2D camera and using a curtain, and a depth camera is utilized to replace manual interaction, so that Trimap acquisition is automatically realized, and the precision and the speed are high.

Description

technical field [0001] The invention relates to the technical field of digital image processing, in particular to a method and system for matting images based on depth information. Background technique [0002] To extract the moving foreground from the static background, in theory, it only needs to determine a background image, and then remove the background image in the new image to obtain the foreground image. However, in most cases, there is no such background image. [0003] Among the existing foreground image extraction techniques, there are methods based on background modeling, foreground extraction methods based on green screen, methods based on pedestrian segmentation, and methods of creating masks based on ternary maps (Trimap). Among them, the method based on background modeling, which extracts the moving foreground from the static background, only needs to use the threshold value to subtract the background in the new image to filter out the foreground, but this m...

AI Technical Summary

Problems solved by technology

Among them, the method based on background modeling, which extracts the moving foreground from the static background, only needs to use the threshold value to subtract the background in the new image to filter out the foreground, but this method has high requirements for the threshold value. The threshold is very sensitive, and it is prone to "double shadow" and "hole" phenomena; while the foreground extraction method based on green screen has the same disadvantages as the method based on background modeling, and it needs to build a green screen studio, which costs a lot of money. The manpower is generally not applicable; the method based on pedestrian segmentation is only applicable to known data set categories, but cannot cover any category, and has certain limitations; and the method of creating a mask based on a ternary map (Trimap) , users need to manually select Trimap, which is time-consuming and inefficient

Images