期刊名称:ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN:2194-9042
电子版ISSN:2194-9050
出版年度:2008
卷号:XXXVII Part B7
页码:541-544
出版社:Copernicus Publications
摘要:In the field of aerial/satellite remote sensing, the quality of data generated by remote sensing platforms is increasing at a significant rate and researches in image compression focus in two aspects: one is seeking efficient compression algorithms suit for remote sensing image data characters, the other is analyzing use potential of current compression algorithms in remote sensing through studying on reduction degree of reconstructed image. The paper aims at experimentally evaluating the effects of compression on image supervised classification. JPEG 2000, which is a new image coding system that uses state-of-the-art compression techniques based on wavelet technology, was employed for compression test. The testing images including three kinds of terrain - plain, hill and mountain area, were compressed with different compression ratios. All the reconstructed images were classified using MLC of supervised classification. The classified results using the original images were used as the reference (truth). Kappa coefficient was adopted as assessment measure. From the results, it can be found that there could be a decrease in image quality with compression ratio increase. While, when the compression ratio is smaller than 10:1, the deterioration of classification accuracy is not significant for both compression methods. As to different terrain types, both image compression techniques have different effects on them. That is, the effects of JPEG2000 and SPIHT compression on image classification are scene-dependent. Generally speaking, with the same compression ratio, JPEG2000 has better performance than SPIHT, which means that JPEG2000 has little deterioration on image classification accuracy than SPIHT