HDR Image Encoding Using Luminance Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for encoding high-dynamic-range (HDR) images are inefficient due to the need for multiple encoding schemes and limitations in dynamic range, resulting in large compressed file sizes and suboptimal performance in applications requiring high compression rates.

Innovation Solution

A method that involves obtaining a low-spatial-frequency version of the luminance component, quantizing it, and calculating a differential luminance component, which is then encoded using frequency coefficients, allowing for efficient encoding and decoding with reduced dynamic range and minimal visual loss.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If traditional encoding schemes are used for HDR images, then the dynamic range is limited to 16-20 bits, but the compression rate is insufficient and file sizes are large

Engineering Contradiction:
Improvedynamic range representationVSAvoidcompression rate
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The HDR image encoding is segmented into two separate components: a quantized luminance component representing the low-spatial-frequency base layer, and a differential luminance component representing the high-frequency detail layer. This segmentation allows each component to be encoded with appropriate precision, achieving both high dynamic range representation and improved compression efficiency

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the encoding parameters by using different bit depths for different components: the quantized luminance component uses reduced precision (lower bit depth) while the differential luminance component uses higher precision. This parameter differentiation allows the system to achieve high overall dynamic range (32-48 bits) while maintaining efficient compression by not allocating excessive bits to all components uniformly

Inventive Principle:
Principle #35Parameter changes

2Manufacturing precision

If multiple encoding schemes are used for HDR images, then the dynamic range can be extended, but the device complexity increases

Engineering Contradiction:
Improvedynamic range representationVSAvoidencoding scheme complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent merges the encoding of quantized luminance and differential luminance components into a unified frequency-coefficient-based encoding framework. Both components are transformed into the frequency domain and encoded using the same transform coding mechanism, simplifying the overall encoding architecture while maintaining the ability to represent extended dynamic ranges

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The differential luminance component acts as an intermediary that bridges the gap between the quantized base layer and the original HDR image. By encoding the difference rather than the full signal, the system achieves high dynamic range representation with reduced complexity, as the differential component contains less information and can be encoded more efficiently

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP3085088B1Method and device for encoding a high-dynamic range image
Publication Date: 2020.04.22 INTERDIGITAL VC HOLDINGS INC
  • EP3085088B1 patent drawingFigure 1~2
  • EP3085088B1 patent drawingFigure 3~4
  • EP3085088B1 patent drawingFigure 5~6

AI summary

The present disclosure generally relates to a method and device for encoding an image block, characterized in that it comprises: - obtaining (101) a low-spatial-frequency version (L lf ) of a luminance component of the image block, said the obtained luminance component of the image block; - obtaining a quantized luminance component (L if,Q ) of the image block by quantizing (102) the obtained luminance component (L lf ) of the image block; - obtaining (103) a differential luminance component (Lr) by calculating the difference between the luminance component (L) of the image block and either the quantized luminance component (L if,Q ) of the image block or a decoded version (L if,Q ) of the encoded quantized luminance component of the image block; - encoding (104) the quantized luminance component (L if,Q of the image block using at least one frequency coefficient of a set of frequency coefficients which is usually obtained from a block-based spatial-to-frequency transform for encoding the luminance component of the image block; and - encoding (105) the differential luminance component (L r ) using the remaining frequency coefficients of said set of frequency coefficients.