Thangka Image Inpainting Method Based On Similarity Resarch

Repair algorithm based on the sample block-based image, combined with the similarity of Thangka image blocks, launched a study of Thangka image to inpaint the problem. Similarity of the image block can be divided into two categories:General similarity,pixel values corresponding to the positions between the two image blocks are equal or similar. Symmetric similarity, pixel values corresponding to the symmetrical position between the two image blocks are equal or similar. Firstly, considering the similarity of the image block symmetry, we propose symmetric sample block Thangka image restoration algorithm based on eight directions. It is primarily a three-step cycle:(1) Enter the priority damaged good image, get damaged boundary points have been split, by comparing the highest priority to get a point, to the point at the center of the square side length of a given block of block-oriented cycles to be repaired. (2) Search for the best symmetry sample block, damaged block from the left, upper left, upper, upper right, right, lower right, lower left, and lower blocks of samples up to find the best symmetry. (3) Repair damaged block, according to the rules of symmetry direction where the best sample block, the dissemination of information to its texture and structure of the damaged block. The above three steps until the area has been circulating the damaged area is zero, repair completed. A large number of experiments show that the algorithm can effectively repair Thangka image symmetrical structure damage. Due to the similarity measure of symmetry sample block and damaged eight directions between blocks is more complex Euclidean distance, you need to consume more time, the improved use city block distance measure and improve the algorithm speed repair. Secondly, in order to have a general, taking into account the damaged block symmetric block may appear in any direction, made symmetrical sample block Thangka image restoration algorithm based on an arbitrary direction. Solve the correspondence between it and the damaged block by rotating the sample block. Experimental results show that the algorithm is very good repair. Because the algorithm is more complex, and is looking for from the whole area of the best non-damaged samples, consumes a lot of time, this approximation algorithm is proposed that the algorithm improves the efficiency of repair. By choosing the size of the sample block repair with different pieces of a picture, select the appropriate size of the sample drawn block not only to achieve the best results and can improve the repair speed of the repair. Finally, the design and implementation of Thangka image restoration system is based on a sample piece of symmetry, a large number of test experiments show that the system should be designed to reach the goal.