| We often say’a picture is worth than a thousand words’, so image plays an important role in our life. With the development of network and digital technique, digital image technology is widely used in various electronic products. But in the process of digital image storage and transmission, the digital image characteristics will cause partial damage.And in other areas of the application, in order to get all the information of the image, some repairs are necessary. Therefore, the research of digital image inpainting becomes very important.According to the different characteristics of image restoration theory and method, common image inpainting algorithm can be roughly divided into two kinds:one is based on partial differential equation model of repair, emphasizing on use of the image structure information to repair the image of small area and scratches. Another kind is the restoration of the model based on texture synthesis. The main method of using image texture information to repair the image with larger amount of information missing or damaged.Firstly, this paper introduced the background knowledge of the digital image inpainting, and give a brief description about mathematical theory and basic. Then discussed the more classic repair models in detail, such as BSCB, CDD and TV, and studied the algorithm theory, the theoretical analysis,the simulation results and the results compared through, summarizes the characteristics of each model. Finally the research analysis algorithm Criminisi, based on texture synthesis of the repair algorithm related improvement, including the priority of the original algorithm, the search area, the similarity of function between the repair block and matching block.The original method of calculating the priority, since the two parameters are normalized, leading to ignore the proportion of known pixels, the improved confidence function changed the proportion of known information, adding the data items in to repair the color of the related information, and improve the reliability of the choice of priority. In order to repair block of matching algorithm using the original is to traverse method, this paper introduced a parameter, through the coefficient to determine the search for matching block, effective shorten the matching of the block for time. In the calculation of similarity between the matching block and repair block, the original algorithm are not considering the texture of the regional information, modification of the similarity formula can consider to distance information, and the limited amount of calculation on the more appropriate place, so as to improve the efficiency of repair. |
