Affine Motion Model Reference Frame Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional motion compensation in video encoding based on block matching assumes translational motion, which is inaccurate for complex motions like translation, rotation, and zoom, leading to incomplete removal of inter-frame correlation and increased bit rates for encoding motion information.
Innovation Solution
The introduction of a high-order affine motion model that allows different motion vectors for pixels in an inter-prediction block, enabling more accurate prediction and motion compensation, along with a mapping function to determine pixel values in a reference frame, improving encoding efficiency by selecting appropriate reference frames for prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional block matching with translational motion model is used, then device complexity is reduced, but prediction accuracy deteriorates for complex motions
Solution Approach 1:
The inter-prediction block is divided into multiple sub-blocks, each with its own motion vector. This segmentation allows different regions to be modeled with different motion characteristics, improving prediction accuracy for complex motions while keeping the overall complexity manageable through localized processing.
Solution Approach 2:
Different motion vectors are assigned to different spatial locations within the block, creating local variations in motion compensation. This local quality approach allows the system to adapt to local motion patterns such as rotation and zoom, rather than applying a uniform translational model across the entire block.
2Measurement precision
If high-order affine motion model is introduced, then prediction accuracy is improved, but bit rate increases due to additional motion information encoding
Solution Approach 1:
Instead of applying full affine transformation with all its parameters, the invention uses a simplified approach where only necessary motion vectors for sub-blocks are encoded. This partial action approach achieves improved prediction accuracy without the complete bit rate overhead of a full affine model.
Solution Approach 2:
The motion model parameters are changed from a single translational vector to multiple localized vectors, but the encoding strategy adapts by selectively encoding only the essential parameters needed to achieve the improved prediction, thereby controlling the bit rate increase.
3Productivity
If multiple reference frames are selected for prediction, then encoding efficiency is improved, but device complexity increases
Solution Approach 1:
Reference frames are pre-organized and prepared in advance, with their availability and suitability determined before the actual encoding process. This preliminary action allows the encoder to efficiently select from pre-validated reference frames without adding significant complexity during the main encoding operation.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Embodiments of the present invention provide a reference frame decoding method, including: obtaining a first reference frame, where the first reference frame is a first picture frame on which decoding reconstruction has been performed or a first interpolated picture frame obtained by means of pixel interpolation on the first picture frame; parsing a bitstream to obtain mapping parameters; determining to-be-determined coefficients of a preset mapping function according to the mapping parameters; obtaining, in the first reference frame according to the mapping function whose to-be-determined coefficients have been determined, a first pixel unit having a mapping relationship with a second pixel unit of a second reference frame; and assigning a pixel value of the first pixel unit to the second pixel unit. In the present invention, only a small quantity of mapping parameters is transmitted at a frame level, and selectivity of reference frames is enriched, so that a reference frame is closer to an actual frame, and prediction performed by using the reference frame is more accurate, thereby improving encoding efficiency.