一种基于稀疏表示的交通图像去噪算法

将图像稀疏表示方法引入到交通图像处理中,实现了一种基于K-SVD的正交匹配追踪的交通图像去噪算法.该算法通过奇异值分解,DCT字典进行自适应更新,形成更能表示图像结构的超完备字典.实验结果表明,相对于传统图像增强方法(中值滤波、均值滤波、基于小波滤波)和基于DCT冗余字典的稀疏表示图像增强方法,该算法能更有效地去除交通图像噪声,得到更高的峰值信噪比.

A Traffic Image Denoising Algorithm Based on Sparse Representation

Image sparse representation method is introduced to process traffic image, and a orthogonal matc- hing pursuit traffic image denoising algorithm is proposed based on K-SVD. The algorithm uses singular value decomposition, so that the DCT dictionary is adaptively updated, forming a over-complete dictionary which can represent the image structure more better. The experiments show that, compared to the traditional image de- noising （median filtering, mean filtering and wavelet filtering） and sparse representation image denoising based on DCT redundant dictionary, this algorithm can remove more effectively the traffic image noise throuh image enhancement and obtain higher signal-to-noise ratio of the peak.