The Research On Key Techniques Of ATR-Directed Image Compression

It is a great challenge for military image communication technology to preserve the auto target recognition (ATR) performance in the process of lossy image compression. The key techniques in ATR-directed image compression are studied in this dissertation.How to measure the impact of lossy image compression on ATR performance is a kernel problem in ATR-directed image compression. A novel image compression quality metric based on the contour information is presented in this dissertation. Based on the fuzzy similarity metric between two shapes, the similarity measure between target contour shapes extracted from source image and codec image is defined as contour feature persistence metric (CFPM). Then the extent to which the ATR performance is preserved in lossy image compression is measured. According to theoretical analysis and experiment results, CFPM measures the extent more comprehensively than correlation coefficients.On the basis of the analysis of region-of-interest (ROI) detection algorithm and ROI prioritized compression mechanism, two ATR-directed compression schemes are presented as: based on full image ROI detection, and based on block ROI judgment. Then the dissertation focuses the issue of ATR-directed aerial reconnaissance optical imagery compression.Man-made blob ROI detection is an important component in ATR-directed aerial optical image coding. This dissertation presents a novel man-made blob ROI detection algorithm based on local intensity entropy feature. According to theoretical analysis and experiment results, this algorithm extracts the local feature difference between the natural background and man-made objects efficiently, and detects man-made target regions robustly with complicated natural environment and different kinds of targets. It can be treated as a reliable detection algorithm for image-compression-directed ROI detection.Based on the ATR-directed image compression system structure and the algorithm of man-made blob ROI detection in aerial imagery, an ATR-directed aerial optical image multi-level ROI coding scheme based on JPEG2000 ROI prioritized compression mechanism is presented in this dissertation. According to theoretical analysis and experiment results, the compressing scheme can meet the requirement of ATR-directed image compression.