Research On Remote Sensing Image Real-time Compression Based On Wavelet Transform

The image real time compression is an important question in remote sensing technique, but the general compression methods haven't achieved the best compression result, therefore developing a real time compression method which is suit for remote sensing image are of significant practical and commercial interest. The remote sensing image has high resolution, weak local correlation, and great information quantity, so the compression method not only need high compression ratio, low distortion, but also need fast compression speed and high reliability. The compression method based on wavelet transform to compress the image data is adopted, after compared the general compression methods and coding criterions. First, the theory of wavelet transform was studies, and it was found that the traditional wavelet transform uses convolution, and it makes the transform process complicated, and the results are floating point numbers. Therefore lifting scheme is used in image compression to decompose image. When wavelet transform is applied to image compression, the chosen wavelet base affects the speed of transform and the quality of the reconstructed image. It is very important to research the correlation between the wavelet base characters and image compression. This paper analyzes the correlation through extensive experiments, and presents the principles of wavelet base choice in image compression. The biorthogonal wavelet base (D5/3) was chosen for remote sensing image real-time compression. Based on the classical wavelet coding methods, a new coding algorithm of fast image compression is presented. The improved algorithm modifies the SPIHT algorithm by using two ideas—"minimum threshold"and "minimum exported bit"to reduces the memory requirement, and establishes "the max value table"in order to lower the repeated calculations. The improved SPIHT algorithm drastically reduces both the memory requirement and the time consumption. At last a new fast compression method which is fit for remote sensing image is designed based on the features of decomposed image and the statistical analysis of the sub-images. The method uses different processing methods for the every sub-image compression. The sub-image with low frequency is compressed losslessly by DPCM method, and the improved SPIHT coding method was used to compress the sub-images with high frequency. Through extensive experiments, the results show that the new compression method is more effective than the others. It improves the compression speed by 3 times, and the PSNR of the reconstructed image been advanced more than 2dB, and the max advanced value is 10.10dB. The human's visual effects of the reconstructed image are close to general algorithms'. In addition, the new compression method is simple, and the memory requirement in the operation process is lower, and it is fit for parallel optimizing progress. The compression speed can be advanced by using DSP, and it also can be implement image real time compression with little information loses.