Research On Locomotive Bottom Image Compression Based On Sparse Representation

Currently,the railway industry develops rapidly,and the railway safety is of vital importance to the national economy and the people's livelihood while the inspection of machine car walk department is the key point of the railway safety.At moment,the inspection of machine car walk department is mainly based on the machine vision.Due to the influence of storage capacity and transmission bandwidth,the efficiency of railway inspection can't match the rapid development of the railway industry gradually.Compression method based on sparse representation is used on the locomotive bottom image in this thesis.Compared to the current JPEG compression,our method could improve the compression efficiency and makes great sense to the inspection of locomotive.Based on sparse representation algorithm,this thesis introduces the basic principle of sparse representation in detail,especially the process of sparse decomposition and dictionary training,then introducing the basic principle of quantization.Finally,the locomotive bottom image is taken as the input image,and the complete compression process is used to test the input image.In this process the uniform quantization is used to quantify the sparse sparse representation coefficient.It is found that the compression result is not satisfying,and the sparse representation coefficient is analyzed by numerical distribution and energy distribution.Improved method is proposed.The improved method is mainly in the improvement of quantization.Firstly,the method of homogenization is proposed.According to the principle of equalization of energy,the nonuniform quantization is carried out,But the phenomenon of white block is occurred when the input image is compressed.Then the quantitative method of sparse representation which quantitation is placed in the process of sparse representation is proposed,and limiting the threshold of sparse coefficients,and finally achieves the uniform distribution of energy and overcomes the phenomenon of white block.The experimental results show that the method proposed in this thesis compresses the input image could achieve better results in terms of compression ratio and image quality recovery compared with the original JPEG compression image.The improved image compression method based on sparse representation is of great significance for railway safety inspection.