RGB-D camera depth image recovery method combining color images

Abstract

The invention relates to a RGB-D camera depth image recovery method combining color images, and belongs to the field of depth image recovery. The method comprises that a Zhang's calibration method isadopted to calibrate color and depth cameras, and internal and external parameters of the cameras are acquired; according to the principle of pin-hole imaging and coordinate transformation, coordinatealignment of the color and depth cameras is achieved; color and depth images of a scene are respectively acquired, and a depth threshold method is adopted to carry out binaryzation processing on theimages; a size of a cavity connected domain is judged to determine existence of cavities; an expansion operation is carried out to obtain a cavity neighborhood; depth value variances of pixel points of a cavity domain are calculated, according to the depth value variances, the cavities are divided into shielding cavities and in-plane cavities; color consistency and domain similarity are respectively adopted to recover the two types of cavities; a partial filter method is adopted to filter the recovered images, and noises are removed. The method ensures the depth image recovery, keeps the original information of the depth images as much as possible, and improves the recovery efficiency of the depth images.

Description

technical field [0001] The invention belongs to the field of depth image restoration, and relates to an RGB-D depth image restoration method combining color images. Background technique [0002] With the development of visual technology, more and more types of cameras are used in display scenes, for example: ordinary color cameras are used for face recognition, CCD cameras are used for obstacle detection, Kinect is used for human skeleton tracking, etc. Ordinary two-dimensional planar cameras can obtain two-dimensional information of the scene. For three-dimensional display space, two-dimensional planar cameras obviously cannot meet the needs of real-world applications. Although two or more two-dimensional planar cameras can be used in combination to achieve three-dimensional information acquisition of the scene , but its use is inconvenient and computationally intensive. This situation has been initially improved with the emergence of RGB-D cameras, such as the Kinect rele...

AI Technical Summary

Problems solved by technology

[0004] In view of this, the purpose of the present invention is to provide a RGB-D camera depth image restoration method that combines color images, and to solve the problem that the depth image information acquired by RGB-D has holes and cannot be used directly, and proposes an effective depth image restoration method. Repair method

Images