Compression Algorithm Research Of Hyperspectral Remote Sensing Image

The outstanding characteristic of compared hyperspectral remote sensing image with multispectral image is the improvement of the resolving power. This for utilize remote sensing image carry on goal classify, discern and with importance value of research of following etc. But its enormous data amount and higher data are linked for transmission and memory of hyperspectral remote sensing image to all bring greater difficulty, so have more extensive meanings to that it is compressed effectively.This text passes with the comparative analysis of other images, the characteristic of hyperspectral remote sensing image of abundant research. Prove hyperspectral remote sensing image relatively stronger spectrum dependence and relatively weaker space dependence.At first, adopt the Adaptive Band Selection to compress spectrum dependence of hyperspectral remote sensing image. The experimental result proves that chooses the validity of the method in adaptive wave band, can choose the wave band with better characteristic in method, it reduce calculation amount that follow-up deal with and improve follow-up punish the result to contribute to.And then, a compression algorithm of Hyperspectral remote sensing image based on Lifting scheme is developed, this method can guarantee the treated image includes the effective information as many as possible. We adopt the splitting of rectangular grids into quincunx grids to construct the second generation wavelets transformation and thereby realize spatial compression. Algorithm introduce one predict with newer two-dimentional little wave the second generation of constructing algorithms because of rectangle bar dose and plum blossom shape bar dose, its Juche idea resolves simple many resolution ratios to the primitive picture , then bar dose and plum blossom shape bar use antithesis promote (predict ) , and primitive to promote (upgrade ) and improvehis performance on the dose in rectangle alternatively, (promote) to a certain characteristic is approached gradually. Promoting the algorithm not only has characteristic such as being simple , fast of structures , should but also be superior to the little wave of first generation to vary on the result of compressing actually, have made the good compression result. Then the decomposition and absolute reconstruction are achieved. The experimental results show that the second generation wavelets based on Lifting scheme can achieve better spatial compression effect than the first generation wavelets.At last, a method combining a dimensional reduction algorithm of adaptive band selection with algorithm based on SOFM codebook design to compress hyperspectral remote images is developed. Because the neural network has fault-tolerant the stronger one, can solve vector atypical match problem of vector in quantizing, and improved SOFM algorithm raise yards of training speed and performance of book. This algorithm has better compression results with utilizing traditional LBG algorithm to design and compare one yard of books. The experimental results show that the SOFM algorithm based on neural network and improving algorithm have a good effect on space compression, which carry out effective compression to hyperspectral remote sensing images.