Abstract: A coding strategy efficiently codes intra (macroblocks, regions, pictures, VOP) data. This strategy uses two basis approaches, a Fixed Deinterleaved Transform Coding approach, and an Adaptive Deinterleaved Transform Coding approach. Furthermore, within each approach, two types of coders are developed. One coder operates on an entire picture or VOPs and the other coder operates on small local regions. Using coders and decoders of the present invention, efficient coding at a range of complexities becomes possible, allowing suitable tradeoffs for a variety of applications.

Abstract: A video encoding method and apparatus is shown wherein image information is represented as a plurality of pixels, the pixels are organized into blocks, pixels transposition is performed on image information at the boundaries of the blocks, the blocks are transform coded and quantized. Pixel transposition involves transposition of alternate pixels at the boundaries of blocks with pixels of neighboring blocks found in a pre-determined direction. The pre-determined direction may be fixed by a system or may be applied on an image by image basis. In the event that the pre-determined direction is not established by a system, a pixel transposition circuit includes a transposition keyword in the output bit stream which is used by a decoded to determine the direction of pixel transposition.

Abstract: A predictive video coder performs gradient prediction based on previous blocks of image data. For a new block of image data, the prediction determines a horizontal gradient and a vertical gradient from a block diagonally above the new block (vertically above a previous horizontally adjacent block). Based on these gradients, the encoder predicts image information based on image information of either the horizontally adjacent block or a block vertically adjacent to the new block. The encoder determines a residual that is transmitted in an output bitstream. The decoder performs the identical gradient prediction and predicts image information without need for overhead information. The decoder computes the actual information based on the predicted information and the residual from the bitstream.

Abstract: Method and apparatus for generating variable length coded for video encoding. A plurality of memory tables contain variable length codes tailored to patterns of input video information that is likely to be input to the variable length encoder. The patterns are determined in large part by a value of a quantizer used in video encoding. A separate memory table containing separate variable length codes is established for each of a plurality of quantization parameter values. The quantization parameter is input to the encoder to enable one of the memory tables and the video information is input to the enabled table to generate the variable length code.