Digital Image Inpainting Technology And Its Application In Image Compression

With the continuous development of social technologies, computer vision and digital image processing has penetrated into every aspect of people’s lives and work. Digital image inpainting is an important research direction in the field of computer vision and image processing, proposed firstly in an academic conference in2000, and now has become a hot research topic for many researchers. Digital image inpainting techniques are widely used, including image compression, repair damaged relics and calligraphy, remove target objects, film and television stunt.The digital image inpainting means to repair the partial region with defect information applying the known region in accordance with certain rules, as naturally as possible after repairing the transition between the defect area and the original known region and minimize manual repair traces, so that the image looks natural.This paper first introduces the definition of image inpainting techniques, significance, and domestic and international development status, and demonstrates the potential applications of this technology. Then the two main image inpainting techniques are studied in this paper:image inpainting technique based on partial differential equation (PDE) and image inpainting technique based on texture synthesis. Applications of the partial differential equations as a mathematical analysis tool in image processing are introduced firstly, and three classic inpainting models on that basis are listed:the BSCB model, TV model, as well as CDD model, including the realization principle of the three models and experimental results.Next the paper introduces image inpainting based on texture synthesis. Firstly, the basic theory knowledge of the texture synthesis technology, and then outlines the existing texture synthesis technology algorithms. On the basis Criminisi et al proposed a sample block-based image inpainting algorithm, which is the key content in the paper. First detailed steps of the classical Criminisi algorithm are studied, and then the advantages of this algorithm:simple and easy to implement, and to consider both the texture information of the image and structure information. An improved algorithm is proposed, which make corresponding improvements for the inadequacies of Criminisi algorithm. The improved algorithm redefine the priority as well as the matching criteria by adding curvature information of the pixels, and set the logical identity matrix to identify whether the sample blocks from the known area is used or not, the sample blocks having been used once are no longer included in the search range. The experiments show that the improved image inpainting algorithm based on the sample block not only in time faster than Criminisi algorithm, but also achieved better results in the quality of image inpainting. Finally, the paper introduces the application of digital image inpainting techniques in image compression. And an image compression system are given integrated the image inpainting technique.First the paper analyzes the feasibility of application of image inpainting techniques in image compression system, and then designs an image codec system by combining existing image compression techniques and improved sample block-based image inpainting algorithm. On the basis of analyzing principle framework of this system, the functional modules of the system and its implementation procedure also are analyzed in detail. In the encoding side of this system, edge of original image is extracted firstly, and then expanded, and then the original edge expanding image is obtained. In the edge expanding image, the target blocks easy to be repaired from the flat region will be discarded, and then the processed image will be encoded and decoded, and finally the edge expanding image after decoding, will be inpainted under the help of the auxiliary edge information to complete the reconstruction of the imperfect image. Experimental results show that image compression system integrated the inpainting technique designed by this paper, under the premise of ensuring the quality of the image, can effectively improve the compression ratio of the image.