LOSSLESS VIDEO CODING SYSTEMS AND METHODS

In some embodiments, a lossless 2D video decoder includes multiple parallel speculative decoding engines, each performing a sample-determination calculation according to one of several possible context values for the current pixel to be decoded. The actual context depends on a quantized value of a local inter-pixel gradient, which may become available only when an immediately-neighboring pixel has been decoded. A multiplexer selects one of the decoding engine outputs when the immediately-neighboring pixel has been decoded and the actual context becomes available. A sample (e.g. luma and/or chroma) value for the current pixel may be determined according to: a baseline prediction generated by applying a median filter to immediately-neighboring pixels; a context-dependent adjustment dependent on local context statistics; and a delta (difference) value read from a video bitstream. Determining multiple context-dependent adjustments in parallel before the actual context is determined allows decoding one pixel per clock cycle.