Digital Image Inpainting And Segmentation

The image repair is an important content in the research of image restoration, is also a hot spot in the research of current image processing and computer vision. The image repair is the process of filling information into the damaged image regions, the goal is restores the damaged image and causes the observers can not detect image defects have been repaired. This technology used in the Heritage's conservation,the film and television stunt manufacture,the old picture's repair,remove texture from image and other objects as well as hide the errors of video, has the very high value of application. This domain's research, the overseas vigorous development, domestic was still being at the start stage. This article mainly studies the small criterion damage image's repair (inpainting) based on the tensor voting algorithm and research the great criterion damage image's repair technology (completion) based on the texture synthesis technology.The first chapter is the introduction, introduced this article's research background,the domestic and foreign research present situation and the application prospect, as well as this article structure and innovation spots.The second chapter introduced the digital image repair's elementary knowledge's commonly used algorithm, introduced the image repair methodology, the inpainting algorithm, the texture synthesis technology, the component labeling algorithm and so on.The third chapter introduced tensor voting algorithm principle, tensor voting in graph restructuring application, tensor voting in image repair application and so on.The fourth chapter introduced the research of the small criterion breakage image repair technology, the small criterion breakage's image repair might regard as is an elimination noise process, this article unifies the tensor voting algorithm and the component labeling algorithm, proposed one kind of new elimination noise's algorithm, compared with the traditional value filter as well as the pure use tensor voting algorithm, regardless of before were processes in the effect or the processing speed surpasses both.The fifth chapter introduced the research of the great criterion breakage image repair technology:The natural image's texture has the very strong directivity. When based on sample texture synthesis technology, search texture match block, the use texture's directional characteristic, may restrain the search process carries on along the texture direction. This chapter in the Criminisi.A algorithm foundation, has joined the determination grain direction process, optimized the texture block priority and the size computational method. The massive experiments indicated: This method when processes the strong directive texture image the repair has the very good effect, distinct enhancement counting yield.The sixth chapter introduced the paper's main innovation achievement and the research conclusion, and proposed some question need the deep research.