Depth image restoration algorithm

Abstract

The invention provides a depth image restoration algorithm. According to the algorithm, morphological filtering is combined with a bilateral filtering algorithm to process a depth image, basic algorithms in morphology are used to fill cavities in the depth image, and bilateral filtering is used to smooth the image. Different types of noise signals included in a target object are eliminated by utilizing different structural elements. The algorithm comprises the following steps that 1) expansion, corrosion, opening operation and closing operation formulas of the morphological algorithm are established; 2) a morphological filter is established; 3) structural elements are selected; and 4) a bilateral filtering criterion is established.

Description

technical field [0001] The invention relates to the field of depth image processing and image restoration. Background technique [0002] Depth images are widely used in the fields of human-computer interaction, 3D scene reconstruction, navigation and positioning; due to the influence of the equipment itself and the acquisition environment during the acquisition process of depth images, there are a large number of holes in the acquired images, resulting in It is difficult to extract accurate depth information. [0003] Many researchers at home and abroad have also conducted a lot of research on the optimization and repair of depth images, such as filling and repairing holes through Gaussian filtering, repairing depth images through bilateral filtering, and repairing depth images through non-local mean Filtering adds edge information weights to upsampling the depth image to repair the depth image, etc., but all methods have defects that can easily cause edge blur, loss of edg...

AI Technical Summary

Problems solved by technology

[0002] Depth images are widely used in the fields of human-computer interaction, 3D scene reconstruction, navigation and positioning; due to the influence of the equipment itself and the acquisition environment during the acquisition process of depth images, there are a large number of holes in the acquired images, resulting in Difficult to extract accurate depth information

[0003] Many researchers at home and abroad have also conducted a lot of research on the optimization and repair of depth images, such as filling and repairing holes through Gaussian filtering, repairing depth images through bilateral filtering, and repairing depth images through non-local mean Filtering adds edge information weights to upsample the depth image to repair the depth image, etc., but all methods have defects, which can easily cause edge blur, loss of edge information, regionality, and large-area holes. The effect is not very good

Images