Image Inpainting Algoirthm Based On Golden Section Pirority Of Block-match

Digital Image Inpainting refers to mending the damaged image area or the wanteddeleted part according to some rules and make it accord with visual psychology ofhuman being, and gaining the satisfied result of inpainting by using the knowninformation of Digital Image. The technology is an important part of Computer Visionand image processing, and it is used in such area as mending of old picture and ancientcultural relics, deleting of words in image, moving of target object, the film visual effectfield.At present, there are mainly two kinds of image restoration method based onstructure and texture. The former is mainly aimed at the images that have obviousstructure information, and is suitable for the unrepaired images with small damagedarea. The latter is mainly for the image that is abundant in texture information, issuitable for the unrepaired images with larger damaged area. This article makes adetailed theoretical and experimental analysis and comparison through five kinds oftypical repair model algorithm. Those are BSCB model based on the structure method,TV model, CDD model, texture synthesis repair model based on texture method,sample block repair model. Then it summarizes the applicable range, advantages anddisadvantages of all kinds of algorithm. There are some questions happened in sampleblock repair algorithm, such as, there will be "fault" phenomenon in the linear structureof area under repair with the damaged area in unrepaired images increase ceaselessly,and a lot of "garbage" will appear during the repair process, visual effect is not good.Directed at those questions, another image inpainting algorithm is designed calledgolden section priority of block-match that is based on the sample block repairalgorithm.The new algorithm firstly making use of image gradient automatically selects thetemplate block size in the repair process, it will choose a larger plate to repair if the image is relatively smooth, and will select a smaller plate to repair if the information ofstructure and texture is complex. Then it will compute the priority of the being fixedpixel blocks using the golden section, using the method of giving the image structural amandatory priority, as well as considering the amount of known information around thepoint of the under repair image, so that it will accord with human visual psychology.Finally, it adds a penalty term to the search for the best matching blocks in the originalalgorithm, pays attention to the reused times of the same block of pixel blocks at thesame time of choosing block match, reduces the repair results with anthropogenic repairvisually. In the filling process it will calculate the average value of the short of the pixelblocks with the best repair priority and the best matching block using the new standardsfor calculating, and fill the average value into the corresponding block of the image, soas to play a compensation role. Experimental results show that this design of therestoration algorithm corresponds with human visual psychology more than the originalalgorithm and has a better visual effect.