Abstract: A digital image preprocessing method for eliminating background patterns and scenes from digitized gray level document image data. The method accepts multi-gray level image data as input and produces a gray level output image containing a "white" background. Document image backgrounds are eliminated (or suppressed) by computing a separate (and typically different) threshold value for each gray level pixel of the original document image. Input image pixels whose gray level is greater (brighter) than the threshold value are assumed to be background picture data and are modified to be "white" pixels in the output image. The pixel threshold value is determined dynamically pixel-to-pixel based on gray level pixel data observed within the input image. Additionally, the pixel threshold can be automatically controlled to insure retention of written and/or printed information present in the input image.

Abstract: A method and apparatus featuring the compression and decompression of the video image data on a single application specific integrated circuit (ASIC) which utilizes typical Huffman coding techniques, but which defines a run count generally as a plurality of zeros between each pair of ones and which further uses a prediction window having substantially no directional biasing errors associated therewith. Additionally, the compression/decompression apparatus of this invention utilizes a reordering technique whereby all of the "high confidence" predictions are placed in one end of the buffer and all the "low confidence" predictions are placed in another end of the same buffer. This reordering, together with the unbiased window configuration and associated run length count definition, allows for high efficiency lossless compression and decompression of associated image data.