Research On Digital Mural Virtual Restoration Technology

The Chinese nation has a history of five thousand years civilization, and there remains a largenumber of precious paintings. Due to the threat of natural disasters and man-made effect,some important murals have suffered serious damage. This paper mainly studies on digitalmural image repair technology, which is the process of Information filling and repair to theinformation defect area on the mural according to the known image information of the mural.Its purpose is to make the observer cannot find out that the mural image had been defected.This technology has a broad application prospect, is a research hotspot in the field of muralrestoration. This technology has a broad application prospect, and had been researched widelyin the mural restoration areas.Firstly, the research background and the Overseas and Domestic Research Status of thedigital mural image was introduced. Secondly two methods of digital mural image patch thatare relatively popular in the academic circles were stated.Finally, the paper optimized the Criminisi algorithm, which is one of the methods basedon the texture information of digital mural image repair. In this paper, the main work has thefollowing three points:(1) Improve the target priority calculation method of the Criminisi algorithm. TheCalculation method multiplies the data item D(p) and the confidence level item C(p), thendetermine the repair order according to the size of the priority. As the repair iteration, the C(p)rapidly falling close to0,however the D(p) increasing. Due to the sensitivity of themultiplication of0, the operation will lead to mismatch. Therefore, the paper try to change theCriminisi algorithm into plus the C(p) and D(p), and introduce a Adjust parameter betweenthem.(2) Improve the target block registration method of the Criminisi algorithm. The searchmethod that used in Criminisi algorithm is global search, whose search scope is the knowninformation area of the whole digital mural image. In this way, search error for the matchingblock would made easily, which will affect the subsequent matches, and finally lead tocontinuous matching error. Therefore, the paper suggested that the texture should besegmentated before Digital mural repair, and then searchs the matching among the split blocksand those blocks contain the split block.Here the paper introduce the pixel distance as a parameter that searching the targetblocks and the matching blocks. When matching among the intersection split blocks, make thebest choice according to the pixel distance of target block and matching block and theresponse pixel error sum of squares.(3) Improve the gray level fusion method of the Criminisi algorithm block. TheCriminisi algorithm block gray level fusion method was put forward by Dr.ZhuWei accordingto the block effect problem that occurs widely in Criminisi algorithm when repairing digitalmural image. The paper realizes the gradient transition of the border pixels repair bysuggesting that use the gradient range of the pixels as quality. The experiments show that compared with the original algorithm,the new one has moresuperiority.