The Application Of Image Texture On Infrared Scene Simulation And Face Recognition

Image texture has been extensively researched in computer vision, pattern recognition and image analysis. However, there is very poor study of combination of texture model-ing, texture analysis and their corresponding applications. This article wants to promote a Cross-over study in multi-context which relate to infrared background simulation based on texture modeling, high-fidelity infrared scene simulation by using real data, extracting spectral feature of face image based on analysis of texture appearance, and face recognition by combining spectral feature and texture feature of face image.In chapter 1, we first point out that the texture takes an important role of scene for-mation, and furthermore, the texture can reflect the spectral information of image. Based on this, we carry out research on infrared scene simulation based on texture modeling, and face recognition based on texture feature. Consequently, we summarize the overseas and domestic study progress of infrared scene simulation and face recognition.Up to now, the infrared scene simulation is lacking in scene model. To solve this prob-lem, in chapter 2, we argue that the infrared texture reflects the infrared radiance energy distribution of material surface by analyzing the formation of it, and this distribution can be learn from samples by FRAME model. And then discuss an infrared background simulation method based on FRAME model, which is a Bayes framework, has powerful descriptive capacity, and can be used to gain the infrared radiance distribution by learning numerous natural infrared textures. This method can be used to produce a scene generator by combin-ing infrared prediction and infrared texture synthesis.The infrared application system always need high-fidelity, multi-time infrared image for testing algorithm. In chapter 3, we present an infrared scent simulation method which can be divided into the infrared prediction and the infrared texture synthesis. This method utilizes real infrared data to gain multi-time simulation result by using VEGA software or data interpolation on real data, and gain high-fidelity, multi-time infrared scene image after embedding infrared targets.The texture appearance is largely dependant on image spectral. Based on this, in chap-ter 4, motivated from psychology research, we suggest that there exist spectral feature of face image by analyzing the texture pattern of different spectral face image. Then we discover the configuration of local interest point detected from face image can reflect the difference of image spectrum, after formulate this problem, we gain the face image spectral feature by using Parzen method.The spectral feature of face image reflects the essence of spectrum, but can not be used for face representation. It needs to be imbedded into the appearance based face representa-tion to enhance descriptive capacity. Based on this consideration, in chapter 5, we present a face representation by combining face image spectral feature and LBP texture feature. This method finds weights set of weighted LBP based face descriptor by integrating spectral in-formation density on the sub-region of face image, and gain the weights set with a natural, generative manner.In the end, the author summarizes the research of this work and puts forward several key problems and future works.