Abstract: For the problem of edge blurry caused by linear interpolation during usual image magnification, a comprehensive model is proposed, which adaptively combines linear diffusion with P-M equation by comparing pros and cons of isotropy and anisotropy diffusion models. In the image magnification, the non-smooth region is processed by anisotropy diffusion model and the smooth region by isotropy model. The experiments show that this model keeps edge sharpness, enhances the clearness of image, and significantly improves magnification image’s perception and objective SNR and PSNR.

Key words: image magnification, partial differential equations, isotropy diffusion, anisotropy diffusion, compreh-ensive model

摘要： 针对传统图像放大处理过程中基于线性插值方法通常导致边缘模糊问题，分析了各向同性扩散模型和各向异性扩散模型在图像处理中的优缺点，提出了一种线性扩散和P-M方程自适应结合的图像放大综合模型。该模型对图像非平滑区域采用各向异性扩散模型处理，而平滑区域则采用各向同性扩散模型处理。实验结果表明，该综合模型在保持图像边缘锐度的同时提高了图像的清晰度，能够有效提高放大图像的主观视觉质量和客观SNR及PSNR。

关键词: 图像放大, 偏微分方程, 各向同性扩散, 各向异性扩散, 综合模型