Ill-Exposed Image Correction via Dual Model Classification

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image processing systems lack a flexible correction mechanism for ill-exposed images, particularly failing to effectively handle multiple exposure types such as back-lit, over-exposed, and under-exposed images across different operating conditions like human perception and computer vision.

Innovation Solution

A correction system and method that utilizes two computational models to classify and adjust original images based on lightness distribution, extracting and adjusting perceptual and structural parameters to convert ill-exposed images into well-exposed images, suitable for various exposure types and conditions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a correction mechanism is designed for a specific type of ill-exposed images, then the correction accuracy for that specific type is improved, but the adaptability to other exposure types deteriorates

Engineering Contradiction:
Improvecorrection accuracyVSAvoidadaptability to exposure types
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent implements a universal correction mechanism that can handle multiple exposure types (back-lit, over-exposed, under-exposed) through a single integrated system. The computing device classifies ill-exposed images into different exposure types and applies appropriate correction algorithms for each type, making the system adaptable to various exposure conditions rather than being dedicated to a single type.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent adjusts correction parameters based on the classified exposure type of each image. By changing the correction parameters according to the specific exposure type (BL, OE, or UE), the system maintains high correction accuracy for each type while using a single universal mechanism, thus resolving the contradiction between specialization and adaptability.

Inventive Principle:
Principle #35Parameter changes

2Manufacturing precision

If image correction processing is performed on all captured images, then the quality of ill-exposed images is improved, but the computing resource consumption increases

Engineering Contradiction:
Improveimage qualityVSAvoidcomputing resource consumption
Core Design Contradiction:
Manufacturing precisionVSUse of energy by moving object

Solution Approach 1:

The patent segments the image processing workflow into two stages: first, a lightweight classification stage that quickly identifies whether an image is ill-exposed and determines its exposure type; second, a correction stage that is only applied to classified ill-exposed images. This segmentation avoids performing heavy correction processing on all captured images, thus reducing computing resource consumption while maintaining image quality for those that need correction.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies correction processing only to the necessary subset of ill-exposed images rather than all captured images. By performing partial action (correction only when needed based on classification), the system improves image quality for problematic images while avoiding unnecessary computing resource consumption on already well-exposed images.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentUS20240163570A1Correction system and correction method for ill-exposed images
Publication Date: 2024.05.16 IND TECH RES INST
  • US20240163570A1 patent drawing
  • US20240163570A1 patent drawing
  • US20240163570A1 patent drawing

AI summary

A correction method for ill-exposed (IE) images, comprises the following steps. (1) A series of original images are captured. (2) The original images are classified as a set of first well-exposed (WE) images and IE images by utilizing a first computational model, according to a lightness distribution of each of the original images. The IE images have a plurality of exposure types including a back-lit (BL) type, an over-exposed (OE) type, and an under-exposed (UE) type. (3) The IE images are corrected to obtain a set of second WE images by utilizing a second computational model. A plurality of perceptual parameters and structural parameters of each of the IE images are extracted and then adjusted according to the BL, OE, and UE types respectively. (4) The first WE images and the second WE images are provided as a set of output images.