Research On Remote Sensing Image Lossless Compression Based On Wavelet Transform

Because of the data of remote sensing image is larger and larger,the space redundant of remote sensing image is lower and lower, the details of remote sensing image richer and richer, the pertinence between the adjacent point is higher and higher, the useless data is bigger and bigger, there must be different degrees of limitations in traditional compression method. Compression technology requires the high compression ratio and low distortion. The remote sensing data compression is in favor of saving communication channel and enhancing the rate of information transfer. The data compression is favorable for carrying out the secure communication and enhancing the overall reliability of system.This article introduces the necessity, possibility and status quo of remote sensing image compression coding. We use wavelet transform to compress the image is based on comparing various compressing arithmetic and coding criterion.Compared with the Fourier transform, wavelet analysis used by the wavelet function (wavelet) has no unicity, the wavelet transform has diversity namely. The character of wavelet function exercises a great influence on coding effect when we use it to compress the remote sensing image. Not all of the wavelet functions are suitable for wavelet image compression. In addition, the character of the corresponding wavelet filter and image compression have important relation, it directly affects the speed and efficiency of transform coding. Therefore, the selection of wavelet image compression is an important issue and there is no uniform criteria about how to select wavelet, which requires users have a detailed understanding on the character of wavelet. Through research we can find that biorthogonal wavelet expends part of the orthogonality, but other character is better than orthogonal wavelet. We usually use biorthogonal wavelet in image processing. The larger wavelet disappearance-distance, the bigger compression ratio; the smaller supported width, the lower computational complexity, rapid implementation can be real; the regularity of the reconstructed wavelet is better, the effect of the image is better. The wavelet which has symmetry can reduce distortion of the edge of the reconstructed image. Then chose the proper wavelet through the wavelet's evaluation-criteria(the greater the coding gain, the better the effect of image reconstruction; The measure of wavelet lossless encoding ability, the value is better when it is lower; the ratio of low-frequency energy and total energy is greater, transform image energy concentration is better, the more it will benefit compression; the greater peak signal to noise ratio of reconstructed image and original image , the better image quality of the reconstruction). Then experiments on the selected wavelet and chose the biorthogonal wavelet——(D9 / 7) on remote sensing image compression.Multi-spectral image is a kind of remote sensing images and become a mainstream gradually. Multi-spectral images are different images which be produced on the same geographical area in different wave band by multi-spectral scanner. For pixels on the image of specific band, the gray value show that the average reflectance values of the corresponding feature in the band, picture of different band and same spatial location make up of band vector. The band vector is corresponding with reflectance curve of features spectral, it is a three-dimensional data, compared with the general image, there is single dimensional multi-spectral information, so the volume of data is very large, generally requires several hundreds to several gigabytes, based on the importance of some specific images , these images lossless compression is essential. Lossless compression use statistical data redundancy compress, the process is irreversible, that is to say the original image data can be fully restored without introducing any distortion, the compression rate of lossless compression is theoretic limited by statistical data redundancy, generally is 2:1 to 5:1. On the basis of several studies in the lossless compression coding method and lossless compression relevant international standards, proposed a lossless compression algorithm based on wavelet transform multi-spectral remote sensing images, on the basis of the analysis of conventional multi-spectral image data and spatial characteristics of wavelet transform coefficient, summing up the correlation of the all-band multi-spectral data wavelet transform coefficient spectral structure. The algorithm is based on the plane coding, and the core is based on spectrum structure coefficient that wavelet transform and design of space-related conditions entropy encoder. At the same time the algorithm is to use transform spectrum and spectrum K-L transform two compression modes. The experiments show that the algorithm of conventional multi-spectral image compression achieved good results; compared with the wavelet encoder the algorithm has been markedly improved. When the statistical correlation of multi-spectral images between the spectra is weaker, the advantage of the algorithm is more obvious.