Hybrid filtering-based depth image hole repair method

Abstract

The invention discloses a hybrid filtering-based depth image hole repair method. The method comprises the following steps of obtaining a to-be-repaired depth image; identifying a hole region of the image; calculating priorities of pixels in a hole; putting the pixels with the priorities higher than a threshold into a priority queue for performing filling; after the repair of the hole region is finished, repairing a non-hole region; and for the hole region and the non-hole region, repairing a non-edge region and then repairing an edge region. The priorities are set for all the pixels in the hole and depend on the support degree and credibility of neighborhood pixels to central hole pixels. With continuous filling of unknown pixels, the hole edge is continuously reduced, the priorities of the pixels of the hole edge are updated, and the threshold is adaptively updated, so that the pixel with the highest priority is ensured to be filled preferentially. Texture and structure information isadded in combination with a bilateral filtering principle, so that a relatively good processing effect is achieved for images with complex textures and structures.

Description

technical field [0001] The invention belongs to the technical field of depth map repair in three-dimensional reconstruction, and in particular relates to a method for repairing holes in a depth map based on hybrid filtering. Background technique [0002] The daily life scenes are all three-dimensional scenes, that is, they are all interwoven with a large number of three-dimensional information, and human beings accept these information on the basis of human visual perception and processing. At present, 3D reconstruction mainly converts 2D information into 3D information through stereo vision technology, and extracts data from images to realize 3D reconstruction of 2D objects. At present, there are two mainstream technologies that can reconstruct 3D scenes: one is to use multi-view methods to reconstruct, and use binocular or multi-eye stereo vision to infer the 3D information of the scene or objects in the scene and display; the other is to reconstruct a 3D model on the bas...

AI Technical Summary

Problems solved by technology

However, the algorithm based on joint bilateral filtering and its improved algorithm focus on the improvement and improvement of the calculation of the depth value of the center hole pixel, and there is not much improvement in the overall hole pixel filling strategy.

This will lead to some unnecessary errors. For example, for an unknown central pixel, although the surrounding adjacent pixels have a high similarity with this central pixel, bu

Images