基于滤波导引冗余字典的图像稀疏去噪  

Image Sparse Denoising Based on Filtered Guidance Redundancy Dictionary

作　　者：温凯峰[1] 叶仕通[2] 

机构地区：[1]嘉应学院计算机学院,广东梅州514015 [2]广东工业大学华立学院,广东增城511325

出　　处：《控制工程》2016年第8期1215-1220,共6页Control Engineering of China

基　　金：广东省高等学校学科与专业建设专项资金(2013KJCX0171);广东省自然科学基金项目(S2013010013307)

摘　　要：研究图像噪声去除步骤中存在局部特征信息缺失问题,提出基于滤波导引冗余字典的图像稀疏去噪方法。该去噪方案利用偏差噪声(附加噪声及对应去噪后图像偏差)进行图像稀疏表达,并对偏差噪声特征信息进行提取,以实现图像去噪效果提升。首先,基于滤波导引对图像去噪后仍存在的偏差噪声进行后处理;然后,基于该偏差噪声设计新的字典训练方法,并自适应获得图像处理冗余字典;最后,基于上述字典对偏差噪声图像进行特征纹理提取,并利用滤波导引方法结合特征纹理提取信息以及去噪图像进行图像恢复,实现图像去噪效果。仿真数据表明,所提算法性能要好于所选对比算法,并且具有较好的视觉恢复效果。To study the problem of local feature information missing in the process of image noise removal, an image sparse denoising method based on filter guided redundant dictionary is proposed. The denoising scheme uses the deviation of the noise(additional noise and the corresponding to the noise of the image deviation) for image sparse expression, and extracts the deviation of the noise feature information in order to achieve the image denoising effect. Firstly, the noise of the image is still in the presence of the noise after the image denoising based on the filtering guidance; Then, a new dictionary training method based on the bias noise is designed, and an adaptive image processing redundant dictionary is obtained; Finally, based on the above dictionary, the texture feature of the feature is extracted, and the image is restored by using the method of filtering guidance. The simulation data show that the proposed algorithm is better than the contrast algorithm, and has good visual recovery effects.

关 键 词：稀疏表达 偏差噪声 冗余字典 图像去噪 滤波导引 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]