Abstract :
A new adaptive multispectral image compression technique based on the regions identified is proposed. The algorithm is adaptive in the sense that according to the data type class of the region, appropriate encoding technique is chosen. The image is first segmented by means of Region splitting and merging procedure based on the statistical characteristic of the image. Then class adaptive hotelling transform or Karhunen Loeve transform in the spectral domain and the shape adaptive wavelet transform in the spatial domain are adopted in the image by considering the spatial, spectral and statistical properties which are unique to the multispectral images. After transformation, based on the regions identified, if the region is relatively uniform or smooth, the SPIHT algorithm is adopted. If not, that is, if the region is highly textured in nature, then object based wavelet method is used for compression. Thus the advantages of both SPIHT algorithm and object based wavelet encoding method, both in terms of visual quality and PSNR values, are incorporated in a single compression technique.
Keywords :
Karhunen-Loeve transforms; data compression; image coding; image segmentation; statistical analysis; wavelet transforms; Karhunen Loeve transform; PSNR values; SPIHT algorithm; adaptive encoding algorithm; adaptive multispectral image compression technique; class adaptive hotelling transform; image merging; image segmentation; object based wavelet encoding method; region splitting; shape adaptive wavelet transform; visual quality; Covariance matrix; Image coding; Image segmentation; Karhunen-Loeve transforms; Merging; Multispectral imaging; Remote sensing; Shape; Transform coding; Wavelet transforms; Adaptive encoding technique; Class adaptive KL transform; Shape adaptive WT;
