基于图像块分组的加权编码去噪方法

Weighting encoding denoising method based on image blocks grouping

针对图像混合噪声去除不足的问题,提出一种分组图像块的加权编码方法。首先,从训练图像中利用非局部相似块提取出分组块;然后,用得到的分组块训练非局部自相似先验模型;最后,集成稀疏先验模型和非局部自相似先验模型到正则化项和编码框架中。实验结果表明,提出的方法在重建图像性能上较同类方法有显著提高,获得了更好的图像恢复质量。

It is hard to remove the image mixed noise,so a weighting encoding method based on image blocks grouping isproposed.The nonlocal similarity block in training image is used to extract the grouping blocks.The nonlocal self-similarity priormodel is trained with the obtained grouping blocks.The sparse prior model and nonlocal self-similarity prior model are integratedinto the regularization term and encoding framework.The experimental results show that,in comparison with other similar methods,the proposed method can improve the image reconstruction performance significantly,and obtain better quality of image restoration.