3D Model Generation from 2D Images Using Focus Stacking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Photogrammetry-derived 3D models often suffer from inconsistencies in detail and focus due to varying depth-of-field and lens effects in 2D images, leading to fluctuating sharpness and detail across the model.
Innovation Solution
A system and method that generates a consistently sharp and detailed 3D model by combining in-focus pixels from multiple 2D images captured with different depths-of-field, using focus stacking techniques and metadata analysis to create a fully in-focus composite image, which is then used to construct the 3D model with uniform sharpness and detail.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If photogrammetry is used to generate 3D models from 2D images, then 3D model generation is achieved, but the model inherits inconsistencies in detail, sharpness, and focus from the 2D images
Solution Approach 1:
The patent segments the 2D images into in-focus and out-of-focus regions using depth information and focus metrics. By dividing the image data into distinct focus zones, the system can selectively process and combine only the in-focus portions from multiple images, thereby eliminating the inheritance of out-of-focus inconsistencies in the 3D model generation process.
Solution Approach 2:
The patent performs preliminary focus assessment and region identification on 2D images before 3D model generation. By pre-identifying in-focus regions and preparing composite images with consistent focus characteristics, the system ensures that only high-quality, sharp data is used in subsequent 3D reconstruction, preventing detail inconsistencies from propagating to the final model.
2Area of stationary object
If multiple 2D images with different depths-of-field are captured, then more complete object coverage is achieved, but the resulting 3D model exhibits fluctuating sharpness and detail across different regions
Solution Approach 1:
The patent applies local quality by assigning different weights and selection criteria to different regions of 2D images based on their focus status. In-focus regions are prioritized and selected for 3D model construction, while out-of-focus regions are excluded or given lower priority. This regional differentiation ensures that each part of the 3D model is constructed from the highest quality available data, maintaining uniform sharpness across the entire model despite varying depths-of-field in source images.
Solution Approach 2:
The patent changes the focus parameter by using multiple images captured at different depths-of-field and selectively combining their in-focus regions. By varying the depth-of-field parameter across the image set and then applying focus-based selection criteria, the system synthesizes a composite representation where all regions contribute their sharpest available data, resulting in a 3D model with consistent sharpness throughout.
3Quantity of substance
If lens effects are present in 2D images, then image capture is achieved, but vignetting and distortions cause peripheral pixels to be less detailed and sharp
Solution Approach 1:
The patent merges multiple 2D images captured from different angles and positions to compensate for lens effects. By combining data from multiple perspectives, the system can select in-focus, high-quality pixels from different images to represent the same object regions, thereby overcoming the quality degradation caused by vignetting and distortions in individual images and achieving uniform detail quality across the 3D model.
Data Source
AI summary
Disclosed is a system and associated methods for generating a consistently sharp, detailed, and in-focus three-dimensional (“3D”) model of an object from two-dimensional (“2D”) images that collectively capture all sides of the object with multiple depths-of-field. The system receives a set of 2D images that capture a particular part of the object with different depths-of-field. The system determines a first pixel from a first 2D image and a second pixel from a second 2D image that represent a common point of the object, determines that the first pixel is out of focus based on the first 2D image depth-of-field and that the second pixel is in focus based on the second 2D image depth-of-field, and defines a 3D construct, that represents the common point in a 3D model of the object, using data of the in-focus second pixel instead of data of the out-of-focus first pixel.