Stereo-Based High Depth-of-Field Imaging Without Z-Stacking

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for generating high depth of field images in microscopes and slide scanners require multiple captures at different depths, which are time-consuming and inefficient, and optical structures for focus adjustment are not suitable for image-based focus determination.

Innovation Solution

A method using a deep learning model to simulate blind deconvolution and stereo imaging techniques to generate high depth of field images without the need for multiple captures, by segmenting regions and estimating depths from stereo images.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If z-stacking technique is used to capture multiple images at different focal planes, then high depth of field image can be obtained, but capturing time increases significantly

Engineering Contradiction:
Improvedepth of field qualityVSAvoidcapturing time
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by performing depth estimation and region segmentation on stereo images before the actual high depth of field image generation. The system pre-identifies regions of interest and estimates depth information, which allows the subsequent image generation process to focus only on relevant areas at appropriate depths, significantly reducing the time required compared to traditional z-stacking that requires capturing multiple focal planes sequentially

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent replaces the mechanical z-stacking system with a computational approach using deep learning models. Instead of physically moving the focal plane through multiple positions and capturing images at each position, the system uses stereo images combined with depth estimation algorithms and region segmentation to generate the high depth of field image computationally, eliminating the time-consuming mechanical focusing process

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Manufacturing precision

If optical structure for multiple capturing is provided to change focal plane, then high depth of field image can be obtained, but device complexity increases

Engineering Contradiction:
Improvedepth of field qualityVSAvoidoptical structure complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent replaces complex optical structures with a computational deep learning-based system. Instead of providing mechanical or optical structures for multiple capturing at different focal planes, the system uses stereo images and applies deep learning models for depth estimation and region segmentation to generate high depth of field images, significantly simplifying the device architecture

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent uses stereo images as a copy or alternative representation that contains depth information, eliminating the need for physical optical structures. By processing stereo images through deep learning models, the system generates high depth of field images without requiring actual optical mechanisms for focal plane adjustment

Inventive Principle:
Principle #26Copying

3Measurement precision

If conventional methods are used for focus determination, then focusing can be achieved, but it is not appropriate for image-based focus determination

Engineering Contradiction:
Improvefocus accuracyVSAvoidimage-based focus determination compatibility
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent changes the parameter basis for focus determination from physical/optical measurements to image-based deep learning analysis. The system uses depth estimation models that process image data directly to determine focus information, making it adaptable to various imaging scenarios including stereo images, while maintaining high focus accuracy through learned patterns from training data

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent replaces conventional focus determination methods with a deep learning-based image analysis approach. Instead of using physical sensors or optical measurements, the system uses neural networks to analyze image content and determine depth and focus information, enabling versatile application across different imaging modalities

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach significantly reduces capture time and enables the generation of high depth of field images from stereo images, allowing for focused images of objects at various positions without the need for complex optical structures.

Implementation Method 1

A method using a deep learning model to simulate blind deconvolution and stereo imaging techniques to generate high depth of field images without the need for multiple captures, by segmenting regions and estimating depths from stereo images.

Methodology Applied
Scientific EffectStereo imaging: Parallax

Data Source

PatentEP4116926B1Method and apparatus for generating high depth of field image, and apparatus for training high depth of field image generation model using stereo image
Publication Date: 2025.08.27 VIEWORKS CO LTD
  • EP4116926B1 patent drawingFigure 1
  • EP4116926B1 patent drawingFigure 2
  • EP4116926B1 patent drawingFigure 3A~3B

AI summary

A high depth of field image generating apparatus according to the present disclosure includes a region segmentation unit which segments a region for a stereo image to generate region data, a depth estimating unit which estimates depths for the stereo image to generate depth data, and a high depth of field image generating unit which generates a high depth of field image from the stereo image, the region data, and the depth data.