基于图像整体变分和分数阶奇异性提取的图像恢复模型

分析基于图像整体变分理论所对应的图像恢复整体变分模型的不足：对于纹理丰富的含噪自然图像，在去除噪声的同时，损失了图像中固有的纹理信息。揭示从残差图像中提取具有分数阶导数奇异性图像的可能性，提出用于图像恢复的残差校正整体变分模型。模型提供一种图像的分解与表示方法。实验结果表明，该恢复模型对自然图像的边缘和纹理等细节保持效果大大优于整体变分模型。

Image Restoration Model Based on Total Variation with Fractional Order Singularity Extraction

Although total variation restoration model is us ed t o denoise the nontexture image, it is not sufficient for real nature images since it removes small scale texture and noise. The drawbacks of the total va riation restoration model are analyzed. The possibility of extracting t he texture with fractional order derivatives filter is studied and an improved total variation restoration model combined with fractional order derivatives sin gularities extracting proposed.The hybrid model provides a new tool to image de compositi on and image modeling. Experiments show the model is suitable to denoising for natural image which is rich of textures, and the model is more effective than th e previous model.