Video Coding Block Boundary Filtering for Flat Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional video codecs suffer from blockiness artefacts in smooth picture areas due to coarse quantization of DCT coefficients, particularly in regions with gradual pixel value changes, such as human faces and skies, leading to unsatisfactory decoded video quality.
Innovation Solution
The solution involves omitting spatial prediction and DCT transform operations for selected image areas with flat or gradually changing pixel values, and applying an interpolative process to define a spatial surface with minimal discontinuities, allowing for visually pleasing decoded images even with coarse quantization by using a surface coding mode that interpolates pixel values across block boundaries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If DCT transform and coarse quantization are used for compression, then bit rate is reduced, but blockiness artefacts appear in smooth picture areas
Solution Approach 1:
The picture is divided into multiple blocks for independent processing and compression. Each block undergoes DCT transform and quantization separately, enabling efficient bit rate reduction while isolating artefact generation to specific regions rather than affecting the entire image uniformly.
Solution Approach 2:
Different quantization strengths are applied to different regions of the image. Smooth picture areas receive lighter quantization to preserve continuity and reduce blockiness artefacts, while complex areas tolerate stronger quantization for better compression efficiency.
2Productivity
If DCT transform is applied to reduce correlation in residual, then coding efficiency is improved, but blockiness artefacts are exacerbated in smooth areas
Solution Approach 1:
The patent applies different processing strategies to different regions: DCT transform is applied in areas with sufficient detail where it improves coding efficiency, while smooth areas are handled with alternative methods that preserve continuity and minimize blockiness artefacts.
Solution Approach 2:
The quantization parameters are dynamically adjusted based on local image characteristics. In smooth areas, lighter quantization is applied to reduce blockiness, while in complex areas, stronger quantization is used to achieve better compression without noticeable quality loss.
3Quantity of substance
If quantization strength is increased for compression, then file size is reduced, but image quality deteriorates in smooth areas
Solution Approach 1:
The patent implements region-adaptive quantization where smooth picture areas are assigned lighter quantization strength to preserve image quality and reduce blockiness artefacts, while complex areas receive stronger quantization to achieve overall compression goals without sacrificing critical detail.
Solution Approach 2:
Quantization parameters are dynamically modified based on local image content analysis. The system identifies smooth regions and applies appropriate quantization strength to balance compression efficiency with quality preservation in those specific areas.
Data Source
Figure 1~2
Figure 3
Figure 4a
AI summary
There is disclosed apparatuses, methods and computer program products for for coding and decoding and specifically but not only for coding and decoding of image and video signals. It is determined whether two adjacent blocks of pixels of an image have a flat nature. The result of the determining is used to select a filter among at least a first filter and a second filter for filtering a block boundary between two adjacent blocks of pixels. The second filter is selected when said determining indicates that there are two adjacent blocks of pixels having a flat nature. At least a first reference value and a second reference value are selected for the second filter and used in filtering the block boundary.