Spherical convergence sample block repairing method for stable filling

Abstract

The invention discloses a spherical convergence sample block restoration method for stable filling. The method comprises the following steps: determining a to-be-restored region and initializing a confidence value; Whether the image is located in a structure area or a texture area is distinguished; Determining the priority of a boundary sample block of the to-be-repaired area, and calculating thepriority by adopting a structure priority ratio or determining a filling sequence according to a spherical convergence priority rule; Searching an optimal matching block according to a Manhattan distance-based matching criterion, and filling a missing part corresponding to the current to-be-filled block with the optimal matching block; And finally, updating a confidence coefficient value of the filling edge by adopting a confidence coefficient updating criterion based on the Stirling theory. And repeating the above steps until the to-be-repaired area is repaired. According to the method, a stable filling sequence can be obtained, the consistency of the repaired image structure information and the texture clarity are kept, the repaired image meeting the human eye vision requirement is obtained, and particularly, the image with the complex structure and texture information can be well repaired.

Description

technical field [0001] The present invention relates to the technical field of image repair, in particular to a sample block-based image repair technology, in particular to a spherical convergent sample block repair method for stable filling. Background technique [0002] Digital image restoration is also called image completion or image de-occlusion. Its purpose is to use the known information around the damaged area to fill the damaged area according to certain algorithms or rules, so that the repaired image looks coherent and natural. Image restoration technology is widely used in the fields of ancient cultural relics protection, restoration of old photos and old movies, film and television special effects production, and error concealment in video communication. It is a research hotspot in the field of computer vision and image processing. [0003] At present, image repair algorithms can be divided into two categories: one is suitable for repairing small-scale damage, an...

AI Technical Summary

Problems solved by technology

[0008] However, the repair results of the above algorithms are prone to structural breaks and wrong extensions.

The algorithm in Document 1 is based on tensor estimation and loses information. When repairing images with rich structure and texture information, the stability and robustness are insufficient.

The filling order of the literature 2, 3 algorithm is unstable and the matching criterion is unreasonable

The Laplacian pyramid used in the literature 4 algorithm is not scalable, resulting in a limited search area

Images