Abstract: High-quality video encoding may be implemented using a single-chip multiprocessor system. Video encoding may be parallelized to take advantage of multiple processing elements available on a single-chip multiprocessor system. Task level parallelism may comprise parallelizing encoding tasks, such as motion estimation, compensation, transformation, quantization, deblocking filtering, and the like across multiple processing elements. Data level parallelism may comprise segmenting video frame data into macroblock partitions and slabs adapted to provide data independence between parallel processing elements. Data communications and synchronization features of the single-chip system may be leveraged to provide for data sharing and synchronism between processing elements.
Type:
Grant
Filed:
November 24, 2009
Date of Patent:
October 7, 2014
Assignee:
Broadcast International, Inc.
Inventors:
J. Dean Brederson, Greg Gillis, Sean Heffernan, Daichi Furusaka, Keishi Chikamura, Masato Hori, Di Wu, Lim Boon Shyang
Abstract: Systems and methods for reducing a bit-rate in a multimedia data stream while maintaining high image quality. Two consecutive frames of the multimedia data stream are spatially filtered, temporally filtered, and combined into a pre-processed frame. In one embodiment, the two consecutive frames are consecutive fields of an interlaced data stream that are deinterlaced by the spatial and temporal filtering process. The pre-processed frame is encoded multiple times using different quantization values to produce a plurality of encoded frames. One of the encoded frames is selected based on its image quality and/or size for inclusion in an encoded data stream. In one embodiment, the pre-processed frame undergoes a motion estimation method that performs an integer level search of overlapping locations around a co-located macroblock in a reference frame followed by a sub-pixel level diamond pattern search.
Abstract: Systems and methods disclosed herein create encoder sensitive video using single and/or bidirectional communication links between a video source and an encoding process to pass metadata (e.g., instructions and cues related to the video stream) to an encoder. A video system includes a video source to generate an uncompressed video stream and metadata corresponding to one or more characteristics of the uncompressed video stream. The video source may include, for example, a video camera or video editing equipment. The metadata may be based on a position, state, movement or other condition of the video source. The system also includes a codec communicatively coupled to the video source. The codec receives the uncompressed video stream and compresses it based on the one or more characteristics indicated in the metadata.
Abstract: Systems and methods for reducing a bit-rate in a multimedia data stream while maintaining high image quality. Two consecutive frames of the multimedia data stream are spatially filtered, temporally filtered, and combined into a pre-processed frame. In one embodiment, the two consecutive frames are consecutive fields of an interlaced data stream that are deinterlaced by the spatial and temporal filtering process. The pre-processed frame is encoded multiple times using different quantization values to produce a plurality of encoded frames. One of the encoded frames is selected based on its image quality and/or size for inclusion in an encoded data stream. In one embodiment, the pre-processed frame undergoes a motion estimation method that performs an integer level search of overlapping locations around a co-located macroblock in a reference frame followed by a sub-pixel level diamond pattern search.