Method and apparatus for low bit rate image compression

Abstract

The invention is a method and apparatus for ordering pixels in a matrix of pixels. The matrix is first divided into multiple cells, where each cell is a smaller matrix of pixels. For each cell in the matrix, an ordering is chosen for the pixels in the cell, and each cell is identified by the pixel ordering for the cell and a direction to a next cell in the matrix. A decision is made on how to represent the ordered block, by first computing a number of non-zero pixels in the current block to a number of runs of zero valued pixels in the current block to compute and intra coefficient count. The pixels in the current block are subtracted from the pixels in a second ordered block in a second image to compute an inter block, and an inter coefficient count is computer by adding a number of non-zero pixels in the inter block to a number of runs of zero valued pixels in the inter block. The block is either discarded, retained, or replaced with the inter block depending on a comparison of the inter coefficient count and the intra coefficient count. Once the block decision is made, the pixels in the block are quantized, using a quantization value determined by computing a first intra coefficient count for a corresponding block in a prior frame. A second intra coefficient count is computed for the block in the current frame, and the percentage difference between the intra counts is scaled by a constant to compute a quantization delta. The quantization delta is added to the quantization value for the prior block to determine a quantization value for the current block.