Image Decoding Device Parallel Processing via Super-Slice Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image decoding devices face inefficiencies in processing high-definition images due to the variable size of slices in bit streams, leading to uneven processing times and reduced parallel processing capabilities, especially with standards like H.264 where a picture can have only one slice or slices of varying sizes.
Innovation Solution
An image decoding device that divides the input bit stream into sub-streams containing prediction units, allowing for efficient parallel decoding regardless of slice size, by grouping macroblocks into sub-streams and utilizing multiple decoders to process these sub-streams in a pipeline fashion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a picture includes only one slice or slices of varying sizes, then the device cannot perform parallel processing or the processing speed is limited by the largest slice, but using multiple decoders with slice-based division would enable parallel processing
Solution Approach 1:
The patent divides the picture into multiple regions called super-slices, where each super-slice contains one or more conventional slices. This segmentation allows the bit stream to be divided into multiple sub-streams that can be processed in parallel by multiple decoders, even when the original slice structure doesn't support parallel processing. The stream divider separates the encoded data of different super-slices into different sub-streams, enabling parallel decoding while maintaining the ability to handle variable slice sizes.
2Productivity
If multiple decoders are used to process slices in parallel, then processing speed increases, but when slices have different sizes the processing time is dominated by the largest slice
Solution Approach 1:
By introducing super-slices that group conventional slices together, the patent creates more granular parallel processing units. Each decoder processes one super-slice at a time, and since super-slices can be formed to balance the workload, this reduces the time imbalance caused by varying slice sizes. The stream divider ensures that encoded data from different super-slices is properly separated and assigned to different decoders.
Solution Approach 2:
The patent dynamically assigns slices to super-slices in a way that balances the processing load across multiple decoders. The super-slice structure allows flexible grouping of conventional slices, enabling the system to adapt to varying slice sizes and distribute work more evenly among parallel processors, thereby reducing the dominant effect of the largest slice on overall processing time.
3Manufacturing precision
If high-definition images with larger data amounts are processed, then image quality and definition improve, but the computational capability and power consumption requirements increase significantly
Solution Approach 1:
The patent segments the decoding process into multiple parallel operations by dividing the bit stream into sub-streams corresponding to different super-slices. This allows the computational workload for high-definition images to be distributed across multiple decoders simultaneously, reducing the power consumption and computational burden on any single device while maintaining the ability to process high-quality images.
Data Source
AI summary
An image decoding device for processing an input bit stream containing encoded data obtained by encoding a moving picture using intra-frame prediction, includes a stream divider configured to divide the input bit stream into a plurality of sub-streams, and a plurality of image decoders each configured to decode the corresponding one of the plurality of sub-streams, thereby outputting images. The stream divider divides the input bit stream so that the plurality of sub-streams each contain the encoded data corresponding to one or more prediction units, where macroblocks of the moving picture each include a plurality of the prediction units for the intra-frame prediction.