The Research On Texture Extraction Of Polarization Image Based On Variational Decomposition

With the progress of science and technology, polarization detection technique is becoming more accurate than before. So we collect more and more valuable polarization images. These data of these images obtained largely contribute to the progress of polarization image processing techniques. Most researchers are continuing focus on that how to extract the necessary features from these images data. The researchers set the correct course by sophisticated image processing technology. In this thesis, we mainly focus on the problem of extracting texture from the polarization images. In recent years, with the increase of the types of polarization camera and the improvement of its accuracy, a lot of polarization images are collected. And it is the trend that the use of computer science techniques to analyze the features of polarization image texture. The visibility of the polarization image is not well than the optical image, but it hides many more information than the optical image. Then, compared with the optical image, the polarization image has better brightness of target and the background and higher contrast. On the basis of classic image texture extraction method, we explore a variational decomposition method to extract polarization images texture. In most of the previous classic methods, the extracted texture mostly has a single scale and orientation. The extracted texture will change with the changing the scale and orientation. This is the shortcoming and deficiency of the classic method. A method of polarization image texture extraction is used based on variational decomposition and LBP algorithm. The experimental results show that the performance of the four texture evaluation indicators is more prominent. This method can provide effective complement and extension for the research on texture feature extraction of polarization image.The central theme of this thesis is to extract the texture of polarization images. We do a series of exploration on this. The main work of this thesis is as follows:Firstly, this thesis introduces the Stokes equations description of polarization image, related features of polarization image related features and the calculation of polarization image used in the experiments.Secondly, it starts with introduction about the definition, types and characteristics of the texture. It describes theoretical definition and mathematical definition of texture. With statement the type of texture, it introduces of classical methods of texture analysis which the LBP algorithm is described in detail. And we apply LBP algorithm to the extraction texture of the polarization image and get the results of the experiment.Thirdly, the algorithm of the variational decomposition is introduced, and some characteristics of its mathematical theory are given. The variational decomposition is mainly based on the energy as the index to extract the texture from polarization images. So the extracted texture has a weak point in other indicators.Finally, the texture images from the variational decomposition experiment are extracted texture by the LBP algorithm. This method fuses the two algorithms. That is to extract the texture from the texture. Polarization image texture extraction is a new approach, and better results of the four evaluations.