Independently Decodable Enhancement Layers in Block-Based Scalable Video Coding Based on RoI Base Layer Blocks

摘要

At least one image of a video stream is encoded (and decoded) in a block based scalable video format, the encoded image comprising a base layer and at least one enhancement layer. Relevant blocks are determined (165) in the base layer according to a criterion that depends on the video data. These may include the potential of candidate blocks to be regions of interest (RoIs), such as small blocks or blocks with high motion values (large motion vectors), and may be identified using relevance maps. The enhancement layer is encoded (155) into at least two distinct groups of blocks (which are independently decodable) as a function of information representative of the determined base layer relevant blocks. Spatial access granularity is controlled. Random spatial access is provided. Blocks corresponding to RoIs can be coded without dependency on other blocks whereas blocks not in RoIs can be grouped together and coded differently to improve compression ratio.