Improved K-mean and Integral Wavelet Based Multi-level Clustering Lossless Compression Algorithm for Remotely Sensed Images

In K-mean algorithm, every pixel in super space is required to calculate Euclidean distance for clustering, so it is one time-consuming hard work when there are a great many class centers. Improved K-mean clustering algorithm presented here can save clustering time by making initial division based on previous clustering results, and maintaining the relationship among stable classes during clustering process. Only calculating and comparing distances for any pixel with neighbor centers near to that pixel except those far away from it, accelerates clustering process with more and more classes becoming stable. Clustering lossless compression algorithm can eliminate efficiently the inter-spectral and intra-spectral redundancy at high convergent speed through enhancing intra-class redundancy. At the same time, one new concept, residue redundancy, is also put forward, whose importance in compression is overlooked previously. The combination of multi-level clustering process and initial S+P integer wavelet transformation can not only remove the spatial and structural redundancy, but also delete the residue redundancy, realizing the breakthrough of lossless compression for multi-spectral images. Furthermore, by comparing with other lossless compression algorithms, the parameter analysis of the TM (Landsat Thematic Mapper) images shows that this multilevel clustering compression algorithm is more reasonable and efficient.