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：2000
卷号：XXXIII Part B7(/1-4)
页码：744-751
出版社：Copernicus Publications
摘要：Image classification, either supervised or unsupervised approach, is an image analysis tool to categorise the unidentified pixels in an image to the designed thematic or spectral separable classes respectively in satellite- based remote sensing discipline. However, the unmanageable physical size of satellite image in somehow increases the cost of transmission, processing and storage of those image data. Using image compression techniques, such as JPEG, can no doubt reduce the physical size of image but the image quality is unavoidable to degrade. Most of these lossy compression techniques are mainly designed to exploit human vision system limitations. The degraded quality of the compressed image may not be visible or obvious when examines by human eyes. As computer-based image analysis tools are very sensitive to image quality, small changes in the image content may affect the analysed results. This paper is to evaluate the effects on the image classification using JPEG-compressed SPOT multispectral images with vary compression quality factors. All the compressed images are classified under supervised classification approach of maximum likelihood classifier (MLC) and unsupervised classification approach of ISODATA clustering (ISOCLUS). With the classified result using original (uncompressed) image as the benchmark, the integrated analysis results of the supervised and unsupervised classification indicates the tolerance of deteriorated classification performance against the compression scale. Finally, some recommendations for the optimum factor of JPEG compression for image classification are concluded
