分组训练卷积字典的图像去噪算法

卷积稀疏编码(convolutional sparse coding,CSC)这一全局模型因字典的特殊结构而受到广泛关注,其中卷积字典学习算法(slice-based dictionary learning,S-BCSC)是最为有效的CSC模型优化算法.虽然S-BCSC算法非常有效,但算法在应用中对整幅图像只使用一个固...

Image denoising algorithm based on group training convolution dictionary

Convolutional sparse coding (CSC) is a global model that has received widespread attention due to the special structure of the dictionary. Among them, the convolution dictionary learning algorithm (S-BCSC) proposed by Elad is the most effective CSC model optimization algorithm. Although the S-BCSC algorithm is very effective, the algorithm uses only a fixed-size dictionary for the entire image in application, but this is not conducive to the accurate description of image information. In order to overcome this shortcoming, this paper discusses how to determine the convolution dictionary size according to the image size, and proposes an image denoising algorithm based on grouping training convolution dictionary combined with sparse representation dictionary learning algorithm. The new algorithm first divides the redundant image blocks into three categories according to smoothness, texture, and edge; then determines the size of the convolution dictionary to be trained for each category; and finally completes the dictionary learning and image denoising process according to the S-BCSC algorithm. It can be seen from the experimental results that the proposed algorithm has improved image quality and clarity compared to the original S-BCSC algorithm.