Subword parallel conditional execution in H.264/AVC deblocking filter implementation

The block-based structure of the H.264/AVC video coding standard can be a source of severe blocking artifacts. Filtering the 4times4 block edges efficiently reduces the visibility of these artifacts, and therefore, a deblocking filter has been made a mandatory in-loop coding tool. Due to the high adaptivity of the filter, conditional processing is needed in block edge and pixel levels. This is especially challenging in subword parallel processors which operate on multiple pixels in parallel. This paper discusses different conditional execution mechanisms used in subword parallel processors. Some of these mechanisms are applied to H.264/AVC deblocking filter implementation. If the processor data path is adjusted according to the 4times4 block size, the edge-level adaptation can efficiently be handled with normal conditional branching, whereas subword parallel conditional execution enables seamless pixel-level adaptation.