Research On Image Inpainting Algorithm Based On Texture

Image inpainting technology is a key point in the field of computer vision and image processing.And it is widely used in people’s life and work.The purpose is to restore the information of defect image or removal of the target object,so that the observers are difficult to perceive the image has ever been edited or not.The method is mainly divided into based on partial differential equations of the repair algorithm and inpainting algorithms based on texture,because the former can not effectively repair damaged areas of the image,so image inapainting based on texture gradually become the mainstream algorithm in the field.In this paper,the image inpainting and image texture background knowledge were introduced,in-depth study of several typical image restoration algorithms is presented,and the corresponding simulation results.At the same time,this paper aiming at the accurate Criminisi algorithm in the repair order and affect the result of image repair factors proposed two improved restoration algorithms.Based on the texture of the image restoration method,the main work is as follows:1.Through the study of Criminisi algorithm,explore the existing problems.Using experimental data,and analysis the impact of the confidence and data item to the repair order,and provides the basis for the further research.2.Consider the texture structure of the image,improve the calculation model of the gradient operator in the data item,and introduce the influence factor,make full use of the known information in the image to improve the priority model,so that the calculation results are more accurate and effective.3.A new method of updating confidence is proposed,and the Prewitt operator is introduced to calculate the gradient and reduce the error between the target block and sample block in the process of updating.Experiments prove that the improved Criminisi algorithm is better.Based on the diffusion adaptive image restoration algorithm,do the following work:1.According to a large number of experiments and data analysis,the damaged area and the texture characteristic value and the optimal sample size selection and adaptive selection of the optimal sample block size.The experimental results show that the image of a reular texture repair,can improve the speed and the effectiveness of the repair.2.Two kinds of smoothing diffusion factors are introduced into the partial differential model,it can along with the structural characteristics of the image,according to the specific direction,around the known information smoothing extends to the inpainting domain.Through the experiment choose appropriate weight,replace data item,define a new priority formula and achieved better repair effect.3.In order to avoid the error caused by the rapid decay of the confidence,improve the reliability of the update method,get a more reasonable repair order.The experimental results show that the combination of these improvements,whether it is to repair the damaged area or the remove of large objects,the improved algorithm has a significant improvement in the effectiveness and speed of the repair.