Image Compression Using Adaptive Dead Zone Quantization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image compression methods often control compression quality through bit rate, limiting local regulation and being unsuitable for applications like satellite image transmission where quality varies across image zones without constrained bit rates.

Innovation Solution

An image compression method using a distortion model independent of bit rate, involving decorrelation, block quantization with a uniform scalar quantizer having a dead zone, and adaptive quantization step determination to achieve target quality in each image block.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If image compression quality is controlled via bit rate, then global compression control is achieved, but local quality regulation is lost

Engineering Contradiction:
Improveglobal compression controlVSAvoidlocal quality regulation
Core Design Contradiction:
Ease of operationVSAdaptability or versatility

Solution Approach 1:

The image is divided into multiple blocks, and each block is independently quantized with its own dead zone parameter T. This segmentation enables local quality regulation while maintaining overall compression control, directly resolving the contradiction between global control and local adaptability.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different blocks of the image can have different dead zone parameters T, allowing each region to be compressed with appropriate quality levels according to its importance. This local quality approach enables selective compression where critical regions maintain higher quality while less important regions use higher compression.

Inventive Principle:
Principle #3Local quality

2Manufacturing precision

If complex iterative processes are used for compression, then compression quality can be optimized, but device complexity increases

Engineering Contradiction:
Improvecompression quality optimizationVSAvoidimplementation complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The distortion model is pre-established with closed-form expressions for quantization distortion D(T) and bit rate R(T). This preliminary preparation of mathematical models allows direct computation of optimal parameters without iterative processes, reducing implementation complexity while maintaining optimization capability.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The invention transforms the complex iterative optimization problem into a simple parameter selection problem by deriving closed-form relationships between the dead zone parameter T and both distortion and bit rate. This parameter transformation approach enables quality optimization through direct calculation rather than complex iteration.

Inventive Principle:
Principle #35Parameter changes

3Productivity

If distortion modeling depends on bit rate, then compression efficiency is improved, but quality control without bit rate constraints becomes difficult

Engineering Contradiction:
Improvecompression efficiencyVSAvoidquality control independence from bit rate
Core Design Contradiction:
ProductivityVSAdaptability or versatility

Solution Approach 1:

Instead of modeling distortion as a function of bit rate, the invention inverts the approach by modeling both distortion D(T) and bit rate R(T) as functions of the dead zone parameter T. This inversion enables quality control through distortion minimization independent of bit rate constraints, while still maintaining compression efficiency.

Inventive Principle:
Principle #13The other way round (Inversion)

Data Source

PatentEP3171594B1Method of images compression enabling to obtain a fixed compression quality
Publication Date: 2021.02.17 THALES SA
  • EP3171594B1 patent drawingFigure 1
  • EP3171594B1 patent drawingFigure 2a~3
  • EP3171594B1 patent drawing

AI summary

Image compression method comprising the following steps applied to at least one image (I): • Decouple (101) the image (I) by applying a mathematical transform to it so as to obtain a set of coefficients, • Decompose (102) the image (I) into blocks and, for each block of the image (I), • Quantize (105) said coefficients using a uniform dead-zone scalar quantizer having a half-dead zone of size T and a quantization step Δ, • Encode (106) the quantized coefficients, • The size T of the half-dead zone of the scalar quantizer being determined (103) by minimizing the gap between an estimated quantization distortion D(T), depending at least on said size T, and a target quantization distortion Dc.