IBC Template Matching for Partially Reconstructed Video Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies, such as HEVC and VVC, face challenges in improving coding efficiency for video processing, particularly in intra block copy (IBC) mode, due to limitations in template matching processes and reference block utilization.
Innovation Solution
A method is introduced that modifies the template matching process for IBC candidates by applying a first template matching process with a modified cost factor, allowing for improved coding effectiveness and efficiency by considering both reconstructed and unreconstructed samples in the current picture.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a conventional template matching process is applied to all block vector candidates uniformly, then the processing is simple to implement, but coding efficiency is insufficient due to not accounting for differences between reconstructed and unreconstructed reference blocks
Solution Approach 1:
The patent applies different template matching processes to different block vector candidates based on their reference block reconstruction status. Specifically, a first template matching process is applied to candidates with partially unreconstructed reference blocks, while a second template matching process is applied to candidates with fully reconstructed reference blocks. This local differentiation optimizes coding efficiency by treating different cases appropriately without uniformly increasing complexity across all candidates.
Solution Approach 2:
The patent modifies the template matching cost calculation by introducing a scaling factor for candidates with unreconstructed reference blocks. The cost is adjusted by multiplying with a factor greater than 1.0, which changes the parameter of cost evaluation to account for the lower reliability of unreconstructed samples. This parameter change allows the system to handle diverse reference block states without requiring entirely different matching algorithms.
2Measurement precision
If template matching cost is not modified for candidates with unreconstructed reference blocks, then the calculation is computationally simpler, but coding precision deteriorates due to unreliable unreconstructed samples
Solution Approach 1:
The patent introduces a cost scaling factor (greater than 1.0) that is applied to the template matching cost of block vector candidates whose reference blocks contain unreconstructed samples. This parameter change adjusts the cost calculation to reflect the lower reliability of such candidates, improving measurement precision without requiring complex re-evaluation of the matching process itself.
Solution Approach 2:
Instead of completely re-evaluating or rejecting candidates with unreconstructed reference blocks, the patent applies a partial correction by scaling the cost. This excessive action (scaling by factor > 1.0) ensures that such candidates are appropriately penalized in the selection process without requiring full re-computation or additional complex analysis, balancing precision and computational effort.
Data Source
AI summary
Embodiments of the present disclosure provide a solution for video processing. In a method, for a conversion between a current video block of a video and a bitstream of the video, a first block vector (BV) candidate of the current video block is determined. The first BV candidate is associated with a first reference block comprising at least one reconstructed sample and at least one unreconstructed sample in a current picture. A first template matching process is applied to the first BV candidate based on a first template matching cost of the first BV candidate between a current template and a reference template, the first template matching process being different from a second template matching process for a second BV candidate. The first template matching cost of the first BV candidate is modified based on a factor. The conversion is performed based on the first template matching process.