This paper proposes effective and efficient multispectral image compression targeting at reducing the extent of storage and transmission time, and at the same time retaining a quality of the reconstructed images. The proposed method is relying on deleting a sub band before compressing the multispectral image. The deleted sub-band is based on the entropy value of each band. Discrete wavelet transform is applied to bands of multispectral image having highest entropy value to delete a sub-band among the four sub-bands and the retained subbands are followed by entropy coder for compression. Furthermore, we exploit JPEG2000+principal component analysis (PCA) to more compress the remaining bands. We used multispectral images from NASA website to validate our proposed method. Experimental result on designated dataset reveals that our proposed method improves the reconstructed image quality better than JPEG2000, SPHIT and some other methods in PSNR and SAM metrics.
展开▼
