Research On Image Inpainting Algorithm Based On Texture Feature

Image restoration technology has become a research hotspot in the field of digital image processing,which aims to estimate the characteristics of known information in digital images,according to certain rules to fill damaged image and makes the repaired effect with easonable structure and image texture.The technology is widely used in the old photo repair,painting and calligraphy relics,murals and network data transmission.In this thesis,an image inpainting algorithm based on texture synthesis is studied.In view of the fact that the image inpainting algorithm based on sample block is d ifficult to keep the texture consistency in the process of image restoration,and it is easy to produce the wrong matching texture block,an improved algorithm is proposed.The main research work is in following filed:1.The improved algorithm combines the curvature feature of isophotes,and the data items introduced reflect texture curvature factor to calculate the priority weights.By adding the weight value,the product of the confidence term and the data item in the original algorithm is improved to form a weighted sum.The influence of the change of the weight value on the image restoration is analyzed.The experimental results show that,with the change of the weight value,the image restoration results will have some fluctuation.It shows that for images with different texture structure,the weight of data items is determined by the texture feature of the image.When the texture feature is abundant,increasing the weight of the data item will increase the sensitivity to the detail texture,so as to improve the repair effect.And the Gauss function is used to update the confidence term,so as to avoid the false matching problem caused by the attenuation of the confidence value.By means of the new updating strategy,the value of the confidence value in the process of image restoration is smoothed.It effectively solves the problem that the calculation of the priority is not accurate when the confidence is zero,so as to improve the effect of image inpainting.2.The commonly used image segmentation algor ithm is analysed and combined with image restoration algorithm.Firstly,the image is divided into several key areas,and then the search area is limited to each key region to ensure the consistency of the structure and texture of the region.It effectively solves the problem of texture extension in inpainting.Finally,the research contents of this thesis are summarized,and the research directions and ideas are given.