Deblocking Filter Level Decision Method for Video Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing deblocking filter level decision methods in codecs like AV1 require extensive calculations and data reading, which hinder real-time hardware encoding performance and violate the execution of the hardware encoding pipeline process.
Innovation Solution
A deblocking filter level decision method that uses a filter level decision of a preceding frame as a reference for later frames, reducing the need for massive calculations and avoiding data reading burdens by searching established tables based on quantization parameters to determine deblocking filter levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full searching method is used to choose deblocking filter levels from sixty-four levels, then the optimum filter level can be selected, but the computational load becomes extremely high (64×4 calculations per frame)
Solution Approach 1:
The patent applies partial action by testing only a limited subset of candidate filter levels (e.g., levels 0, 1, 2, and 3) rather than performing exhaustive full searching across all sixty-four levels. This selective approach achieves sufficient filtering quality for most practical scenarios while dramatically reducing computational complexity from 64×4 to just 4×4 calculations per frame, thereby resolving the contradiction between accuracy and encoding speed.
Solution Approach 2:
The patent performs preliminary action by pre-calculating and storing optimal filter level mappings in lookup tables based on quantization parameter ranges before actual encoding. During encoding, the system simply queries these pre-computed tables using the current QP value, avoiding real-time exhaustive searching and enabling fast filter level selection that maintains both quality and speed.
2Measurement precision
If iterative trial method is used to find optimum filter level by observing filter effects, then accurate filter level can be determined, but repeated data reading increases hardware pipeline burden
Solution Approach 1:
The patent performs preliminary action by pre-determining optimal filter levels through offline analysis and storing them in lookup tables indexed by quantization parameter ranges. During hardware encoding, the system simply performs a table query based on the current QP value, eliminating the need for iterative trial filtering and repeated data reads, thus simplifying the hardware pipeline while maintaining accurate filter level selection.
Solution Approach 2:
The patent uses copying by creating pre-computed lookup tables that contain optimal filter level mappings for various QP ranges. Instead of performing complex iterative trials during encoding, the system copies the appropriate pre-determined filter level from the table based on the current QP, significantly reducing computational overhead and hardware pipeline complexity.
3Measurement precision
If deblocking filter level decision is made for each frame independently through full searching, then optimal filter level can be achieved, but the encoding process becomes too slow for real-time applications
Solution Approach 1:
The patent applies partial action by testing only a limited subset of candidate filter levels (e.g., levels 0, 1, 2, and 3) rather than performing exhaustive full searching across all sixty-four levels. This selective approach achieves sufficient filtering quality for most practical scenarios while dramatically reducing computational complexity from 64×4 to just 4×4 calculations per frame, thereby resolving the contradiction between accuracy and encoding speed.
Solution Approach 2:
The patent applies parameter changes by dynamically adjusting the number of candidate filter levels to test based on the quantization parameter value. For higher QP values where blocking artifacts are more prominent, more candidate levels may be tested, while for lower QP values fewer candidates are needed. This adaptive approach optimizes the balance between filtering quality and encoding speed across different video content characteristics.
Data Source
AI summary
Disclosed is a deblocking filter level decision method applicable to an image encoder and used for determining deblocking filter levels of N image frames. The method includes: determining whether a current quantization parameter of a current frame of the N image frames is the same as a previous quantization parameter of any previous frame of the N image frames; if the current quantization parameter is different from the previous quantization parameter, looking up M current deblocking filter reference levels in an established table according to the current quantization parameter and then determining M current deblocking filter levels of the current frame according to the M current deblocking filter reference levels, respectively; and if the current quantization parameter is the same as the previous quantization parameter, determining the M current deblocking filter levels of the current frame according to M previous deblocking filter levels of the previous frame, respectively.