基于像素中位扩散的大面积损坏图像修复机制

摘要

In order to overcome the deficiencies of current image inpainting algorithms such as obvious blur effect and block effect in inpainting large area damaged images with high contrast edges and high frequency components,we proposed the pixels median diffusion-based inpainting mechanism for large area damaged image.We induced anisotropic diffusion model to eliminate the interference of noise on the illumination line of the inpainting boundary of damaged image;and constructed the slope limitation criterion of pixels gradient mode to improve the stability of image inpainting mechanism;then we embedded weighted factor,and designed the pixels median diffusion-based image inpainting model according to local maximum likelihood estimation mechanism,and completed image inpainting.Finally,with the help of simulation platform,we tested the performance of the mechanism.Results showed that comparing with other image inpainting algorithms, when inpainting large area damaged image and the image with high contrast edges,this mechanism had better inpainting effect,higher similarity,and effectively reduced the blur effect and block effect.%针对当前的图像修复算法在修复高对比度边缘和高频分量的大面积损坏图像时，存在明显的模糊效应与块效应等不足，提出基于像素中位扩散的大面积损坏图像修复机制。引入各向异性扩散模型，消除噪声对损坏图像修复边界等照度线的干扰；构造像素梯度模的斜率限制准则，以提高机制稳定性；嵌入加权因子，基于局部最大似然估计机制，设计基于像素中位扩散的图像修复模型，完成图像修复。最后借助仿真平台，测试了该机制性能。结果显示：与其他图像修复算法相比，在大面积损坏图像与高对比度边缘图像修复中，该机制具有更好的修复效果和更高的相似度，有效降低了模糊效应与块效应。