Block-Shape Transform Selection for Lower-Bitrate Video Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding systems face inefficiencies in selecting optimal transform modes for video blocks, leading to suboptimal compression and increased bitrate, particularly in hybrid video codecs like H.265/HEVC and VVC, due to the fixed use of transform types without considering the shape of the transform block.
Innovation Solution
A method and apparatus for adaptively selecting a transform mode based on the shape of the transform block, using discrete cosine transforms (DCT) and discrete sine transforms (DST) tailored to the block's dimensions, such as DCT-2 and DST-7, to optimize compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a fixed transform type is used for all video blocks, then the device complexity is reduced, but the compression efficiency deteriorates
Solution Approach 1:
The patent implements dynamic transform mode selection where the transform type (DST-7 or DCT-2) is adaptively chosen based on the actual shape of each transform block. The encoder determines whether the block is taller than wide and selects the appropriate transform mode accordingly, allowing the system to adapt to varying block shapes rather than using a fixed transform type for all blocks.
Solution Approach 2:
The patent applies different transform modes to different regions (blocks) based on their local characteristics (shape). Each transform block is evaluated individually, and the transform mode is selected locally according to whether the block is taller or wider, rather than applying a uniform transform mode across the entire video stream.
2Productivity
If transform mode selection is adapted to block shape, then compression efficiency is improved, but device complexity increases
Solution Approach 1:
The patent implements a simplified adaptive approach where only the block shape (taller than wide or not) is considered for transform mode selection, rather than analyzing all possible block characteristics. This partial adaptation provides compression efficiency improvement while keeping the complexity increase minimal by focusing on the most significant shape characteristic.
Solution Approach 2:
The patent changes the transform mode parameter (DST-7 or DCT-2) based on the block shape parameter. By linking the transform mode selection to a simple geometric parameter (block aspect ratio), the system achieves adaptive compression efficiency improvement while avoiding the need for complex analysis of multiple block characteristics.
3Quantity of substance
If DST-7 transform is used for taller blocks, then bitrate is reduced, but manufacturing precision requirements increase
Solution Approach 1:
The patent changes the transform parameter (selecting DST-7 vs DCT-2) based on the block shape parameter. This parameter-based selection allows the system to optimize bitrate for different block shapes while maintaining implementation simplicity by using well-defined transform standards with fixed mathematical formulations.
Data Source
AI summary
An apparatus includes at least one processor; and at least one non-transitory memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: determine a horizontal dimension of a transform block corresponding to a horizontal direction of the transform block and a vertical dimension of the transform block corresponding to a vertical direction of the transform block; determine a first transform and a second transform, wherein the first transform has smaller first coefficient for its first basis function compared to a first coefficient of a first basis function of the second transform; select the first transform having smaller first coefficient for its first basis function for the direction with the smaller dimension of the transform block and select the second transform having larger first coefficient for its first basis function for the direction with the larger dimension of the transform block.