带有高阶耦合项的各向异性扩散图像复原

Anisotropic diffusion with high-order coupling term for image restoration

二阶各向异性扩散进行图像去噪容易产生块效应。针对这个方面的不足,通过在整体方差模型中引入高阶耦合项,使图像边缘得到保持的同时得到分片光滑的结果。分析了耦合系数和迭代次数对图像复原结果的影响。实验结果证明,引入高阶耦合项的新模型能够很好地防止块效应的产生。新模型与其他两种模型(整体方差模型、四阶偏微分复原模型)的图像复原相比,最大峰值信噪比比整体方差高出2dB,比四阶偏微分高出1dB左右。

Image denoising with second order anisotropic diffusion often leads to blocky effect.After adding high-order coupling term to total variance model,the new model produced piecewise smooth results,while preserving sharp jump discontinuities in images.The influence of coupling coefficient and iterations on restoration effect were analyzed.Experimental results show that the new model alleviates the blocky effect.The method achieves at least 2dB gain over total variance denoising model,and about 1dB gain over four-order PDE denoising model for max peak signal-noise rate （max PSNR） measurement.