Adaptive Transform Domain Filtering for Video Compression

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video encoding standards, such as H.264, face challenges in achieving optimal rate-distortion performance due to lossy quantization noise, which is not effectively addressed by existing post-processing filters applied in the pixel domain.

Innovation Solution

Implementing adaptive filtering in the transform domain, specifically using discrete cosine transform (DCT) coefficients to reduce quantization distortion before reconstruction, by generating and transmitting filter coefficients in the bitstream for encoding and decoding processes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If post-processing filters are applied in the pixel domain to remove quantization noise, then distortion is reduced, but computational complexity increases

Engineering Contradiction:
Improvereconstruction qualityVSAvoidcomputational complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent applies filtering in the transform domain (frequency domain) rather than the pixel domain (spatial domain). By transforming the filtering operation from spatial to frequency domain, the patent reduces computational complexity while maintaining reconstruction quality, as filtering in the transform domain can be performed more efficiently on the already-transformed coefficients.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Manufacturing precision

If adaptive filtering is applied in the transform domain to reduce quantization distortion, then rate-distortion performance is improved, but overhead increases due to filter coefficient transmission

Engineering Contradiction:
Improverate-distortion performanceVSAvoidbitstream overhead
Core Design Contradiction:
Manufacturing precisionVSLoss of information

Solution Approach 1:

The patent applies adaptive filtering selectively to different transform coefficients based on their local characteristics. Filter coefficients are generated and transmitted only for specific frequency bands or coefficient groups where quantization distortion is most problematic, rather than applying uniform filtering across all coefficients. This localized approach improves rate-distortion performance while minimizing the overhead from transmitting filter coefficients.

Inventive Principle:
Principle #3Local quality

3Ease of manufacture

If uniform quantization step size is used in DCT coefficient quantization, then encoding simplicity is maintained, but quantization noise and distortion increase

Engineering Contradiction:
Improveencoding simplicityVSAvoidquantization distortion
Core Design Contradiction:
Ease of manufactureVSManufacturing precision

Solution Approach 1:

The patent introduces adaptive filter coefficients that modify the quantization process dynamically. While the base quantization step size remains uniform for simplicity, the patent applies adaptive filtering in the transform domain using coefficients that are adjusted based on local signal characteristics. This allows the system to maintain encoding simplicity while reducing quantization distortion through the adaptive filtering stage.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP2617199B1Methods and devices for data compression with adaptive filtering in the transform domain
Publication Date: 2018.01.24 BLACKBERRY LTD
  • EP2617199B1 patent drawingFigure 1
  • EP2617199B1 patent drawingFigure 2
  • EP2617199B1 patent drawingFigure 3

AI summary

Methods and devices for encoding and decoding data using adaptive transform domain filtering are described. The encoder determines a set of transform domain filter coefficients to be applied to a transform domain prediction. The filtering may, in some cases, also apply to quantized transform domain coefficients. Rate-distortion optimization may be used to determine the optimal filter coefficients on a frame-based, coding-unit-basis, or other basis.