Sergei V. Vasilyev (SOLERC, Institute for Electromagnetic Researches)
Real-time image recording tasks do require reliable parallel compression algorithms to unload the memory and data transmission channels. A new method based on contouring the image content with subsequent converting of the contours to a compact chained bit-flow, thus providing efficient spatial image compression, is proposed. It can be directly applied to the 1-bit images and have been tested for encoding the electronically published scanned article images. The tests have shown that this technique gives a near-equal compression to the CCITT Group4 2D method, which is, for example, currently used in the ADS electronic article services, but it allows additional 20 to 30 per cent compression through the subsequent arithmetic or Huffman-type coding. Combining this method with pattern recognition and edge detection techniques can provide an adequate tool for automated mapping/compression of color images in remote sensing applications. Possible ways to such combination are discussed. As the method allows operation in the scanning-line mode, it also conjoins well with the learned principal components spectral compression proposed by the author earlier (http://www.stsci.edu/stsci/meetings/adassVII/vasilyevs.html) and mutual implementation of these two techniques may become promisingly efficient for the on-board multispectral astronomical image compression. Details of the contour-based image compression algorithm will be presented.