Gaze-Adaptive Image Compression for Low-Latency VR

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image compression methods are inadequate for low latency applications like virtual or augmented reality, as they require significant processing time and do not achieve sufficient compression, leading to restricted image resolution and quality, especially in mobile solutions.

Innovation Solution

A method and apparatus that selectively encodes frequency coefficients using a bit encoding scheme based on transmission bandwidth, quality of service, and pixel position, discarding higher frequency components and using varying bit numbers to achieve optimal compression and decompression, allowing for efficient image transmission in wearable digital reality headsets.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If traditional image compression methods (JPEG, DCT-based) are used, then bandwidth requirements are reduced, but processing time increases significantly and compression ratio is limited

Engineering Contradiction:
Improvedata volumeVSAvoidprocessing time
Core Design Contradiction:
Quantity of substanceVSLoss of time

Solution Approach 1:

The patent extracts and encodes only the most significant frequency coefficients (luminance and selected chrominance coefficients) while discarding less important ones. This selective extraction achieves high compression ratios by transmitting only essential visual information, reducing data volume without requiring full processing of all image data.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies different encoding precision to different frequency coefficients based on their visual importance. Luminance coefficients are encoded with higher precision than chrominance coefficients, and coefficients in different frequency ranges use different bit allocations. This local differentiation optimizes compression while preserving perceived image quality.

Inventive Principle:
Principle #3Local quality

2Manufacturing precision

If high resolution images are generated to avoid motion sickness, then image quality improves, but processing hardware requirements and bandwidth increase

Engineering Contradiction:
Improveimage resolutionVSAvoidbandwidth
Core Design Contradiction:
Manufacturing precisionVSQuantity of substance

Solution Approach 1:

The patent transforms image data from spatial domain to frequency domain using DCT, then selectively encodes frequency coefficients with varying bit precision. This parameter transformation allows high-resolution imagery to be represented with fewer bits by concentrating visual information in fewer significant coefficients, reducing bandwidth requirements while maintaining perceived image quality.

Inventive Principle:
Principle #35Parameter changes

3Manufacturing precision

If more frequency coefficients are encoded, then image quality is preserved, but compression ratio decreases

Engineering Contradiction:
Improveimage qualityVSAvoiddata volume
Core Design Contradiction:
Manufacturing precisionVSQuantity of substance

Solution Approach 1:

The patent applies different encoding strategies to different regions of the frequency spectrum. Luminance coefficients (which contribute most to perceived image quality) are encoded with higher precision and retained, while chrominance coefficients are encoded with lower precision or discarded. This local quality differentiation maintains essential visual quality while achieving high compression ratios.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent selectively discards less important frequency coefficients (particularly higher-frequency chrominance coefficients) that contribute minimally to perceived image quality. By discarding these coefficients and retaining only the most visually significant ones, the patent achieves substantial compression while preserving essential image quality.

Inventive Principle:
Principle #34Discarding and recovering

4Quantity of substance

If compression is increased to reduce bandwidth, then data volume decreases, but image resolution and quality are restricted

Engineering Contradiction:
Improvedata volumeVSAvoidimage resolution
Core Design Contradiction:
Quantity of substanceVSManufacturing precision

Solution Approach 1:

The patent uses variable bit-depth encoding for different frequency coefficients, allocating more bits to luminance coefficients and fewer bits to chrominance coefficients. This parameter differentiation allows aggressive compression of less important data while preserving resolution and quality of visually critical information, achieving high compression without restricting essential image quality.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3472806B1Image compression method and apparatus
Publication Date: 2025.08.06 IMMERSIVE ROBOTICS PTY LTD
  • EP3472806B1 patent drawingFigure 1
  • EP3472806B1 patent drawingFigure 2A~2B
  • EP3472806B1 patent drawingFigure 3

AI summary

A method of compressing image data from one or more images forming part of digital reality content, the method including obtaining pixel data from the image data, the pixel data representing an array of pixels within the one or more images; determining a position of the array of pixels within the one or more images relative to a defined position, the defined position being at least partially indicative of a point of gaze of the user; and compressing the pixel data at least partially in accordance the determined position so that a degree of compression depends on the determined position of the array of pixels.