Research On Correction Method Of Row Displacement For Passive Millimeter Wave Image And The Fusion Algorithm With Visible Image

A passive millimeter wave (PMMW) imaging system achieves images through thedifference of intensity that scenes and objects radiate, which has great performance inpenetrating and excellent ability of recognizing a metal target from its surroundingenvironment. These features make it ideal for battlefield reconnaissance and securitychecking of important places. However, because of the limits of hardware, the spatialresolution and Signal Noise Ratio (SNR) of PMMW image is relatively low, leading tothe lack of detailed description, which cannot totally meet the application requirement.Image fusion is the combination of information from different sensors, which cancombine the intuitive optical image with PMMW image, and achieve thecomplementary of advantages and the improvement of image quality.Based on the actually specific research projects, this article studies thedisplacement estimation and correction method to solve the dislocation problem thatoccurred in the scanning process. On the other hand, this article gives an image fusionalgorithm based on Pulse Coupled Neural Network (PCNN) and second generationCurvelet. The mainly work are as follows:(1) The theories and methods of PMMW imaging technology and image fusion areintroduced and analyzed. The features of PMMW image are summarized, whichprovides the direction and emphasis of following research.(2) The continuity of adjacent rows of data is proposed when the sensor imaging toa natural scene. Correlation value is introduced to measure this continuity. In thefrequency domain, the similarity of low-frequency and the difference of high-frequencyare analyzed.(3) The reason of the dislocation problem of PMMW image is analyzed. Thedislocation is estimated based on the correlation of two adjacent rows and thephase-difference function. By using the linear phase shift compensation in the frequencydomain, the displacement in the spatial domain can be corrected, which can achieveaccuracy of sub-pixel level without interpolation. In the actual project, the effectiveness and practicality of the proposed method has been demonstrated, which can solve thedislocation problem of the PMMW image in the actual application.(4) The Pulse Coupled Neural Network (PCNN) theory is introduced, and thefeatures when it’s applied in image processing are analyzed. Aiming at the problem thatPCNN model has lots of parameters, an adaptive linking strength selection method isdesigned. The PCNN model is improved aiming at the PMMW image processing, andprovide an important theories and methods for the subsequent image fusion strategyselection.(5) Aiming at the fusion of optical image and PMMW image, this article gives animage fusion algorithm based on the second generation Curvelet and imporved PCNN.Region growing is used for the PMMW image segmentation, leading to the discard ofnoise and useless information in the fused image. The simulation results show that theproposed algorithm can achieve the fused image which is more in line with human’svisual habits, and be able to achieve a better result in both subjective and objectiveevaluation criteria.