Adaptive Filter Parameter Derivation for Video Loop Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing video coding systems face challenges in reducing processing delay and buffer requirements due to the sequential and data-dependent nature of loop filtering processes, particularly with adaptive filters like SAO and ALF, which increase system cost and latency.
Innovation Solution
The proposed method involves deriving adaptive filter parameters from pre-in-loop video data, allowing concurrent application of in-loop and adaptive filters, and using a moving window approach to process image units, reducing the need for large buffers and processing latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If adaptive filters (SAO and ALF) are applied sequentially after in-loop filtering, then picture quality is enhanced, but processing latency and buffer requirements increase
Solution Approach 1:
The patent performs adaptive filter parameter derivation in advance using pre-in-loop filtered video data. By deriving parameters before the in-loop filtering is complete, the system prepares filtering parameters preliminarily, allowing the adaptive filters to start processing as soon as sufficient filtered data is available, rather than waiting for complete in-loop processing
Solution Approach 2:
The patent divides the video data processing into segments using a sliding window approach. Instead of processing entire frames sequentially, it processes smaller windowed portions of video data, allowing parallel processing of different segments and reducing the overall processing latency while maintaining picture quality
2Manufacturing precision
If adaptive filters are applied sequentially after in-loop filtering, then filtering accuracy is improved, but buffer size requirements increase
Solution Approach 1:
Parameter derivation is performed preliminarily using available pre-filtered data, so that when adaptive filtering needs to occur, the parameters are already prepared. This eliminates the need to store large buffers of pre-filtered data waiting for parameter completion
Solution Approach 2:
By using sliding window segmentation, the patent reduces buffer requirements from frame-level to window-level buffering. Each window only requires local neighboring data for parameter derivation and filtering, dramatically reducing the stationary buffer size needed while maintaining filtering accuracy through local adaptivity
3Manufacturing precision
If filter parameters are derived from in-loop processed video data, then filtering performance is improved, but memory bandwidth demands increase
Solution Approach 1:
Parameter derivation is performed preliminarily using pre-in-loop data that is already in memory, avoiding the need to re-access original unfiltered data during adaptive filtering. This preliminary parameter preparation reduces subsequent memory bandwidth requirements while maintaining filtering performance
Solution Approach 2:
The sliding window approach segments memory access to only the necessary local regions required for parameter derivation and filtering. Instead of accessing entire frames, the system accesses only the current window and its immediate neighbors, significantly reducing memory bandwidth demands while preserving filtering performance through local adaptivity
Data Source
AI summary
A method and apparatus for loop processing of reconstructed video in an encoder system are disclosed. The loop processing comprises an in-loop filter and one or more adaptive filters. The filter parameters for the adaptive filter are derived from the pre-in-loop video data so that the adaptive filter processing can be applied to the in-loop processed video data without the need of waiting for completion of the in-loop filter processing for a picture or an image unit. In another embodiment, two adaptive filters derive their respective adaptive filter parameters based on the same pre-in-loop video data. In yet another embodiment, a moving window is used for image-unit-based coding system incorporating in-loop filter and one or more adaptive filters. The in-loop filter and the adaptive filter are applied to a moving window of pre-in-loop video data comprising one or more sub-regions from corresponding one or more image units.