Research On Image Inpaiinting Based On The Sparsity Of Image Structure

The digital image inpainting is important branch in the domain of image processing. The goal of image inpainting is to repair the damaged area of the image with the known area according to the algorithm or the rules. The image inpainting can be divided into two kinds: small scale regional image inpainting and image inpainting based on texture synthesis.The former’s main ideal is to change the image inpainting process into PDES function models, so we can deal with the image with intelligent optimization method. This kind of algorithm establishes the model on the boundary of the damaged image, and then fills the damaged area automatically by smoothly propagating imformation from the surrounding areas along the normal of direction isophote. It also smooths the noise while preserving the edges.The latter makes use of the consistency of the texture region for the efficient texture synthesis. It is mainly suitable for the texture image with big damaged area. The Criminisi algorithm laid the foundation of image inpainting based on texture synthesis. Firstly, find the block with the biggest priority on the boundary of the damaged image, then search the known region for seeking the most similar block for the current block and substitute the current block. Finally, updata the boundary and repeat the above process. The advantages of the Criminisi algorithm are faster and efficient. The Xu’ algorithm provides a new priority with the structure sparse. It has a bigger promotion than Criminisi algorithm. But the two algorithms are instable in the priority, and the repair order easily lead to the wrong matching. This paper presents a thorough study of the existing algorithms, and has made two improvements to its existed insufficiencies:First, the priority improved. Based on the existing priority, we improve a new priority with the geometric distribution. Firstly, we calculate the discrete of the similarity block, and then calculate the priority of the block. It has more reasonable repair order.Second, priority repair strategy and regional division. Through the analysis and research of the visual theory of image, we put forward a new repair strategy of structure priority.When repair the structure, we fitting the edge to determine when to stop the structure repair. Finally, we recover the smooth region and texture region. Compared with the existing image inpainting algorithm, our algorithm has a bigger improvement in repair the edge of damaged image.At the same time, based on the theoretical analysis, this article carries on a larger number of simulation experiments of the proposed improved algorithm to verity the feasibility and effectiveness of the algorithm.