Video Image Decoding With Cross-Frame Reference-Block Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding technologies introduce large prediction errors in motion information, leading to low encoding efficiency due to the increasing proportion of motion vector information in the code stream.
Innovation Solution
The method involves determining reference blocks using known motion information and symmetry searching to reduce prediction errors by calculating sums of absolute differences, and optionally incorporating template similarity comparisons to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion information is obtained through template matching with reconstruction values, then motion vectors can be determined, but large prediction errors are introduced resulting in low encoding efficiency
Solution Approach 1:
The patent performs motion estimation between reference frames before template matching to obtain preliminary motion vectors. These preliminary motion vectors are then used to guide the template matching process, reducing the search space and improving accuracy. This preliminary action prevents large prediction errors while maintaining encoding efficiency.
Solution Approach 2:
The patent introduces an intermediate step of motion estimation between reference frames that acts as a mediator between the available reference data and the final motion vectors. This intermediary process generates more accurate prediction information that reduces prediction errors in the subsequent template matching stage.
2Loss of information
If motion vectors are obtained by template matching in reference frames, then motion information can be derived, but the proportion of motion vector information in the code stream increases requiring effective compression
Solution Approach 1:
The patent extracts only the essential motion information from the template matching process by using motion estimation between reference frames to obtain preliminary motion vectors. This extraction approach reduces the amount of motion vector data that needs to be encoded while preserving the most important motion characteristics, thereby reducing code stream size.
Solution Approach 2:
The patent discards redundant motion information by relying on motion estimation between reference frames for the bulk of motion prediction, and only recovers essential motion details through selective template matching. This approach reduces the quantity of motion vector information in the code stream while maintaining necessary prediction accuracy.
3Ease of manufacture
If the current block is a partition block with unknown reconstruction values of neighboring blocks, then template construction is incomplete, but using predicted values introduces additional prediction errors
Solution Approach 1:
The patent performs motion estimation between reference frames as a preliminary action to obtain motion vectors before attempting template construction. These preliminary motion vectors provide the necessary information to construct templates even when reconstruction values are unavailable, reducing the need to use predicted values and thereby reducing additional prediction errors.
Solution Approach 2:
The patent enables the decoding process to be self-sufficient by using motion estimation between reference frames to generate the necessary motion information without relying on potentially inaccurate predicted values from neighboring blocks. This self-service approach improves prediction accuracy by eliminating the error propagation from using predicted values.
Data Source
Figure 1~3
Figure 4
Figure 5
AI summary
A method, an apparatus and a device for obtaining motion information of video images are provided. The method includes: determining at least one first candidate reference block corresponding to a current block in a first reference frame according to known motion information related to the current block; determining at least one second candidate reference block that is respectively corresponding to the at least one first candidate reference block in a second reference frame; calculating a sum of differences between a first candidate reference block and a second candidate reference block corresponding to the first candidate reference block; and selecting the first candidate reference block and the second candidate reference block which have the minimal sum of differences as a first reference block and a second reference block of the current block, and obtaining motion information of the current block according to motion information determined by the first reference block and the second reference block. A template construction method is also provided. The method, apparatus and device can improve the encoding and decoding efficiency of video images.