Research Of Stain Inpainting In Car Video Based On The Improved Criminisi Algorithm

With the development of digital processes, digital image and video has been a major carrier for information exchange. Digital image and video processing technologies has made a great development and achievement. Image inpainting and video inpainting are the technologies which use known information from the known region to fill in the missing region. Therefore, we can get the image which looks like it never been damaged.Car video can record the environment outside the vehicle. But during the process of grabbing the images by the camera, there would be some stains on the camera. Then the images will be damaged by the stains. In this paper, solid stains can be detected and repaired so that the image may looks like it never been damaged.(1)In the first paragraph, the background and the research situation at home and abroad has been illustrated. Otherwise, the definition of image inpainting and video inpainting and related visual theory has been introduced. Some classic algorithms has been analyzed and compared.(2)In the process of grabbing images, there will be some noise. In this paper, the method of median filtering has been used to eliminate the noise. In addition, the image segmentation method based on Fisher function was improved so that the image can be segmented by an adaptive threshold. A new segmentation method was proposed, it used the information from the time domain, using the differential method to eliminate the moving objects and remain the stain. The final testing results can be got after the morphological processing.(3)For the reason that the stain is usually a large black area, the Criminisi algorithm was selected to remove the stain. Considering the character of the stain, the algorithm has been improved. In order to ensure the integrity of the structure, the priority function has been changed by adding the weight of the structure. The local search method replaced the global search method in order to improve the efficiency and the speed. In the process of matching the target block and the matching block, the matching result was more accurate with the geometrical distance being a parameter. The experimental results showed that the improved algorithm can be more accurate and faster.(4)The last part was the inclusion of the research. The direction of the future research has also been illustrated. It will be a hope that the method in this paper can be used in more areas, and there will be a more effective method to solve the problem of stain removal.