Research And Implementation Of Infrared Image Preprocessing System Based On Retinex

Due to the influence of materials of the infrared detector, imaging mechanism and the environment, infrared thermal images tend to have the characteristics of narrow dynamic range, low contrast, high noise and fuzzy details, which increases the difficulty for the image observation, feature extraction and target recognition, etc. The paper carries out the research on image enhancement and real-time implementation by combining the advantages of Retinex algorithm with infrared image characteristics.Compared with the traditional enhancement algorithms, Retinex algorithm has more advantages in the aspects of dynamic range compression, color fidelity and detail enhancement, etc. So in this paper, a deep research on Retinex is carried out and an Retinex image enhancement algorithm of non-subsampled wave transform based on the guided filter is proposed, which focuses on the infrared image with low gray value, low contrast and high noise. With the combination of the multi-scale of non-subsampled wavelet transform and the borderline of guided image filtering, the Retinex image enhancement algorithm of non-subsampled wave transform based on the guided filter is implemented. Different from the traditional Retinex algorithm based on some kernel functions, the guided image filtering can bring in the input image, which can adjust output weight values adaptively according to adjacent pixels mean values and variance values to realize the image enhancement.The infrared image pre-processing system based on FPGA is designed in this paper. Verilog HDL language is used to design the communication function module. The system achieves the goal of image receiving, pre-processing and display. Considering the timing characteristics and the real-time requirements of pretreatment system, we put forward and implement a new kind of Retinex image enhancement algorithm that is suitable for the FPGA timing requirements.Based on the infrared image preprocessing system, the experiments of the Retinex image enhancement processing are implemented, the results show that the enhanced image has increased dynamic range, enhanced contrast, and meets the requirement of real-time processing.