Tile Boundary Noise Reduction in HEVC Video Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The video coding method in the H.265 (HEVC) standard introduces discontinuous noise at tile boundaries, which degrades image quality due to the use of tiles for enhancing parallel processing capability, and existing de-blocking filters do not sufficiently reduce this noise.
Innovation Solution
A video encoder with dual motion vector detection units processes video signals from adjacent tiles to generate motion vectors that preferentially reference images from other tiles, reducing noise at tile boundaries through inter prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If tiles are introduced to enhance parallel processing capability, then productivity is improved, but discontinuous noise is generated at tile boundaries degrading image quality
Solution Approach 1:
The patent applies different motion vector detection strategies for different regions: for blocks at tile boundaries, it preferentially selects reference blocks from adjacent tiles, while for blocks away from boundaries, conventional intra-tile reference selection is used. This local differentiation resolves the contradiction by maintaining parallel processing benefits while specifically addressing boundary noise through localized reference selection modification.
Solution Approach 2:
The patent introduces an intermediate reference block selection mechanism that acts as a mediator between tile boundaries. By selecting reference blocks from adjacent tiles for boundary blocks, it creates a smooth transition zone that reduces discontinuous noise while preserving the overall tile-based parallel processing structure.
2Object-affected harmful factors
If existing de-blocking filters are used to reduce noise at tile boundaries, then image quality is improved, but the noise reduction is insufficient
Solution Approach 1:
The patent performs motion vector optimization at the tile boundary before de-blocking filter application. By pre-selecting optimal reference blocks from adjacent tiles for boundary blocks, it reduces the magnitude of discontinuous noise beforehand, making the subsequent de-blocking filter more effective and reducing the need for aggressive filtering that could degrade image quality.
Solution Approach 2:
The patent implements a feedback mechanism where the motion vector detection unit identifies blocks adjacent to tile boundaries and automatically adjusts reference block selection based on this detected condition. This feedback loop ensures that boundary blocks use cross-tile references while internal blocks use conventional references, optimizing noise reduction without compromising overall image quality.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
To reduce noise or the like generated at a boundary of tiles introduced in a video coding method. In a motion vector detection unit, a first tile video signal and a second tile video signal included in one picture are supplied to a first detection unit and a second detection unit, and a reference image is supplied from a frame memory to the first detection unit and the second detection unit. The first detection unit performs processing, by inter prediction, on the video signal positioned on or in the vicinity of a tile boundary between a first tile and another tile among many video signals included in the first tile. In this processing, the first detection unit generates a motion vector so as to preferentially refer to the reference image included in another tile different from the first tile among the reference images read out from the frame memory.