Image Classification for Automatic White Balance

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional Automatic White Balance (AWB) technologies distort image colors when applied to images with insufficient color change or monochromatic objects/backgrounds, as they fail to accurately differentiate between monochromatic and non-monochromatic images.

Innovation Solution

An image classification apparatus and method that divides input images into blocks, calculates hue and chroma values, determines color-changed blocks, and uses hue and chroma variances to classify images as monochromatic or non-monochromatic, applying AWB only to non-monochromatic images with significant color changes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If AWB is applied to images with insufficient color change or monochromatic objects, then white balance adjustment is performed, but color distortion occurs in the image

Engineering Contradiction:
Improvewhite balance adjustment accuracyVSAvoidcolor distortion
Core Design Contradiction:
ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent applies preliminary action by performing image classification before AWB processing. The system divides the image into blocks, calculates hue and chroma values, and determines whether the image is monochromatic or non-monochromatic before applying AWB. This preliminary classification prevents color distortion by avoiding AWB application to unsuitable images while ensuring proper white balance adjustment for appropriate images.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If conventional AWB methods use average RGB values or color difference signals for gain adjustment, then white balance is recovered for images with sufficient color change, but color distortion occurs when applied to monochromatic images

Engineering Contradiction:
Improvecolor measurement accuracyVSAvoidadaptability to different image types
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent applies local quality by analyzing local color characteristics through block division. Instead of treating the entire image uniformly, the system divides the image into multiple blocks and calculates hue and chroma values for each block. This local analysis enables the system to accurately identify monochromatic regions and make informed decisions about AWB application, improving both measurement precision and adaptability to different image types.

Inventive Principle:
Principle #3Local quality

3Productivity

If AWB is applied based on insufficient color change detection, then processing is performed on images that may not require it, but the complexity of accurate classification increases

Engineering Contradiction:
Improveimage processing efficiencyVSAvoidclassification system complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent applies segmentation by dividing the image into multiple blocks and analyzing color characteristics at the block level. This segmentation approach enables efficient classification by processing smaller regions independently, calculating hue and chroma values for each block, and determining monochromaticity based on local variations. This method balances processing efficiency with classification accuracy, avoiding excessive system complexity.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP2222088B1Apparatus and method for classifying images
Publication Date: 2019.08.14 SAMSUNG ELECTRONICS CO LTD
  • EP2222088B1 patent drawingFigure 1
  • EP2222088B1 patent drawingFigure 2
  • EP2222088B1 patent drawingFigure 3

AI summary

An image classification apparatus and method for Automatic White Balance (AWB) are provided. An input image is divided into blocks including pluralities of pixels. A hue value and a chroma value are calculated for each of the blocks. A color-changed block is detected by calculating, for each one of the blocks, differences between the hue and chroma values calculated for all blocks adjacent to the one of the blocks and the hue and chroma values calculated for the one of the blocks. A hue variance and a chroma variance are calculated for the entire input image if a number of the detected color-changed blocks is greater than or equal to a first threshold. The input image is determined as a non-monochromatic image, if the hue variance is greater than or equal to a second threshold or the chroma variance is greater than or equal to a third threshold.