Based On The Automatic Identification And Classification Of The Region Of Interest For Remote Sensing Image Compression Technology

Remote sensing image have been playing an important role increasingly. With the increase of the resolving power of them, the amount of remote sensing image is larger and larger. They both cause that the conflict, between limited channel capacity, limited memory capacity and the exigent need of transmission and disposal of plenty of remote sensing image, is more and more outstanding. Therefore, researches on compression are more and more important. In order to resolve, to a certain extent, the conflict between compression ratio and reconstructed image quality in current compression, combining the fact that ROI in a great number of image varies with different users, researches have been performed. The researches are based on city objects, as ROI, and on visible light remote sensing images to bring forward a novel compression method. In it the reconstructed ROI quality is as high as the need of users and the reconstructed BG quality is as low as the need of users.Firstly, feature space is gotten through analysis of the city composed of different objects. In the space, with the method of risk evaluation, several values are induccd to distinguish ROI from BG according to sample images and features extracted according to disposal images. Split-combination based on the result of comparing the feature to the value and structure modifying are in turn carried out to get recognition result. Secondly, respective compression level is gained according to users and classifying compression is performed. At last, reconstructed image is gotten through orderly decoding ROI and BGThe research is realized by c language, at the same time, a hardware system based on ICETEK-DM642-PCI is realized. The result of above both systems testifies that the way is realizable and advanced. The result of test through many remote sensing images holding city zone with various figure is as follows: The way can auto-recognise city zone in visible light remote sensing images with the high recognition precision, and can conveniently classifying compress ROI and BG according to users and the compression performance is better than those of the compression with JPEG or JPEG2000 solely. The superiority is that the compression ratio of the way is higher than those of compression with JPEG or JPEG2000 solely on condition of same ROI psnr and is that the ROI psnr of the way is higher than those of compression with JPEG or JPEG2000 solely on condition of same compression ratio.