Digital Image Inpainting Algorithm Research

Digital image inpainting technology is an important branch in the field of image processing, computer vision research hot spots. The main use is made of the existing image information technology, according to certain algorithm or rules to the damaged area for repair or removal of the target, in order to achieve image has good visual effect after repair or close to the purpose of the original image visual effect.Digital image inpainting technology research mainly focused on three aspects:the inpainting algorithm based on partial differential equation, the inpainting algorithm based on texture synthesis and repair algorithm based on sparse representation. This thesis introduces the research significance and current situation of the development of digital image inpainting technology, as well as the theoretical basis of the algorithm.Based on the study and summary on the basis of predecessors’ fruitful research results, this thesis proposed to the contain both small pieces of information loss and contain large image inpainting technology of lack of useful information and iterative shrinkage algorithm in image inpainting, the main contribution of this dissertation can be summarized as follows:1) In image inpainting, for processing contain small pieces of missing information containing large pieces of information missing image algorithms rarely involved, although Criminisi and MCA algorithm can inpainting this kind of image, but the algorithm fixed visual effect is bad and time-consuming longer, so the CDD is combined with the advantages of the proposed model and Criminisi algorithm to inpainting damaged area difference, namely the CDD model combined with Criminis algorithm.2) Before the image processing, histogram equalization is used to strengthen details, on the one hand, to expand the dynamic range of the image, guarantee for the continuity of the restoration image structure; On the other hand, restrain the noise in the image, make the image visual effect is better after repair.3) Based on the existing iterative shrinkage algorithm based on the analysis of the research, put forward a kind of image inpainting algorithm based on parallel coordinate descent method, each iteration through the linear search to determine the drop step length, then determined by threshold shrinkage drop direction, to ensure that the algorithm is to inpainting the image with high quality.