基于压缩感知的道路交通图像处理及重构算法研究

Road Traffic Image Processing and Reconstruction Algorithms Based on Compressed Sensing

在现代道路交通图像处理的研究中,由于传统图像数据压缩采样受到带宽限制,导致重构图像数据含的有用信息量偏少.为此,提出一种压缩感知的道路交通图像处理方法,提高了重要信息含量且大大减少了交通图像数据存储空间.首先,在采样率为0.6的情况下,在dbN系列小波基变换下以分段正交匹配追踪(StOMP)算法对原图像进行重构仿真,研究了不同阶数消失矩的小波稀疏表示与重构图像性能的关系.其次,研究了不同采样率对图像重构性能的影响,包括重构时间、重构误差、峰值信噪比.最后,用BP,OMP,StOMP算法对图像重构进行研究,为道路交通图像数据存储和图像重构提供了理论基础.

In the study of modern road traffic image processing, the traditional compression sampling of image data is limited by the bandwidth, leading to the reconstructed image data containing few useful information. To this end, the road traffic image processing method is presented based on the compressed sensing to improve the important information content and to greatly reduce the traffic image data storage space. First of all, when the condition of the sampling rate is 0.6, dbN series wavelet transform and stagewise orthogonal matching pursuit （STOMP） algorithm are used to study the relationship between wavelet sparse representation and image reconstruction of dif- ferent vanishing moments. Secondly, the influence of different sampling rates on the performance of image reconstruction is studied, including reconstruction time, reconstruction error and peak signal to noise ratio. Finally,BP,OMP and STOMP algorithms are used to study the image reconstruction,which provides a theoretical basis to the road traffic image data storage and the image reconstruction.