关键词: 多光谱图像压缩, 空间稀疏等价表示, 聚类, 主成分分析, 小波编码, 误差补偿, 码率分配

Abstract:

Aiming at the problems of inadequate reducing of spatial and spectral redundancy and weak adaptability of existing multispectral image compression algorithms, the multispectral images' spatial sparse equivalent representation and its clustering implementation named the OptimalLeaders are proposed. Furthermore, an adaptive multispectral image compression algorithm—OLPKWS is designed, which is based on composite transform. In the OLPKWS, multispectral data are transformed into representation and residual by the presented spatial sparse equivalent transform, which removes spatial redundancy adaptively. Moreover, an error compensation mechanism is introduced in order to improve the quality of the reconstruction image. Principal Component Analysis (PCA) is used to remove spectral redundancy for representation. However, to predict differences, KLT is utilized to explore spectral correlation, two-dimensional wavelet transform is used to remove spatial redundancy, standard deviation weighted rate allocation and SPIHT are combined to complete the coding. Experimental results show that, in comparison to the clustering, SPIHT and KLT_SPIHT_TCIRA algorithms, the proposed approach achieves a higher peak signal to noise ratio (PSNR) under the same compression ratio.

Key words: multispectral image compression, spatial sparse equivalent representation, clustering, principal component analysis, wavelet coding, error compensation, rate allocation