Sparse Decomposition For Traffic Image Compression

With the rapid development of economic construction, the urban traffic problem is getting more and more serious, and the modern management is an urgent need to manage the transportation. Therefore intelligent traffic system is studied. In intelligent traffic system, image information as important information has been always collected and used because it is intuitive and rich in content. However, it has already become one of the major factors of blocking intelligent traffic system from application extensive, because it can't meet the requirement of real-time transmission in the large and high-quality traffic image.In the recent years, sparse decomposition becomes very popular in the study of image processing. It is one kind of non-orthogonal decomposition, decompose the image on the over-complete dictionary. The decomposition result is very concise. Because it can transoform an image into a spare formation, it has become a new way of solving the image compression in low bit rate.In order to solve the image compression in intelligent traffic system, the paper introduces sparse decomposition, and according to characteristics of the traffic image, focuses on the traffic image compression issues. First, the article study the similar characteristics of the traffic image background and local, then carry on the sparse decomposition to the traffic image, at last the distribution of the traffic image sparse decomposition result is analyzed and image coding seheme is proposed. The main work and research are as follows.1. The principle of image sparse decomposition and image sparse representation are introduced. To the large computation of image sparse decomposition, fast algorithm based on particle swarm algorithm (PSO) is used.2. To the disadvantages of sorting difference algorithm, background difference compression algorithm of the traffic image is proposed. According to the similar characteristics of the traffic image background, the traffic image is pretreated by background difference compression algorithm, and then is sparsely decomposed. At last, an image coding seheme is proposed based the images sparse decomposition results. Experiments show that the introduced method can release background information redundancy of the traffic image. Compared with the sequence difference algorithm, the algorithm improves not only the PSNR of decoded image, but also improves subjective image quality under the same compression ratio.3. To enhance the traffic image compression effect, a novel forecast and quantization of atomic parameters based on traffic image compression is proposed. First, after sparse decomposition every time, atomic parameters are forecasted according to the similar characteristic of traffic image. The distribution of the atomic parameters of quantization error is analyzed, a coding seheme is proposed based on the atomic parameters of quantization error. The experimental results show that the algorithm effectively reduces the code redundancy of the atom parameters. Compares with the sorting difference algorithm and the background difference compression algorithm of the traffic image, the algorithm improves not only the PSNR of decoded image, but also improves subjective image quality under the same compression ratio.4. In order to effectively describe the traffic images, a layered compression algorithm based on the traffic images is proposed. The algorithm introduces thought with layers into the sparse representation and coding process of the traffic image. According to the similar characteristic of the traffic image, the atomic dictionary of the algorithm is constructed by the decomposed traffic images, and replace the over-complete dictionary. The traffic image is sparsely decomposition by the greedy algorithm. The residual image is coded by the set partition in hierarchical trees. The experimental results show that compared with the forecast and quantization of atomic parameters based on traffic image compression, the set partition in hierarchical trees and JPEG, the algorithm improves not only the PSNR of decoded image, but also no blocking artifacts and improves subjective image quality under the same compression ratio.The paper was supported by the National Natural Science Foundation of China (60702026), the Scientific and Technological Funds for Young Scientists of Sichuan (09ZQ026-040), the Open Research Fund of Key Laboratory of Signal and Information Processing, Xihua University (SZJJ2009-003) and the Open Foundation of Engineering Research Centre of Transportation Safety of the Ministry of Education of China (WHUTERCTS2009A01).