Criminisi image restoration method based on HSV color space

Abstract

The invention provides a Criminisi image restoration method based on HSV color space, comprising the following steps: 1, judging whether there is an image damaged area to be repaired; 2, taking the pixel point p of the damaged edge, calculating the confidence value C according to the pixel value in the square area of 9*9 centered on the point p, judging the size thereof, and if the confidence value C is great than the threshold value K, calculating the priority value P of the square area block according to the priority function; 3, according to that matching principle, the b matching block inthe undamaged region is searched by comparing the difference value of the hue H and the luminance V in the HSV color space; 4, filling the searched best matching block into the pixel block to be repaired, and making the central pixel points coincide; 5, updating that boundary information of the repair area, and simultaneously updating the confidence value of pixel point of the repair completed area; 6, repeating that above steps until all pixel block are repaired. The invention makes up for the defects of inaccuracy of priority function in Criminisi restoration algorithm and degradation of image quality after restoration, and has excellent performance.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a Criminisi image restoration method based on the HSV (Hue, Saturation, Value, hexagonal pyramid model) color space. Background technique [0002] Research on image and restoration has become an important research topic in image analysis fields such as edge detection, image segmentation, machine vision, and pattern recognition. The edge structure and texture information of an image can reflect the basic features and important information of the image content, but it is very easy to cause image damage during image acquisition, compression or transmission, thus affecting subsequent work. Image inpainting technology is based on the estimation of the current image structure information, and reasonably restores the damaged image with partial missing or occluded objects, which has considerable application value. Due to the lack of prior information, the repair problem is ofte...

AI Technical Summary

Problems solved by technology

Due to the lack of prior information, the repair problem is often ill-conditioned, so mathematical methods such as partial differential equations (PDE) need to be used, but for images with large damage, the repair effect is not ideal

The data items in the Criminisi image repair algorithm based on texture synthesis will eventually tend to zero as the damaged area is gradually repaired, resulting in a situation where the confidence value and data items are one large and one small, which will cause errors in the calculation of pixel priority in the later stage , this error will gradually amplify as the restoration proceeds, thus affecting the final restoration effect

[0003]At present, there are many researches on image restoration, but for some internal texture features and edge corner information, the gradient operator cannot fully achieve the ideal restoration. Effect

Images