Depth Map Generation via Blur Difference Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for determining depth maps and autofocusing, such as hill-climbing and focal gradient methods, require controlled lighting, are computationally expensive, or rely on specific scene features, limiting their effectiveness and accuracy.
Innovation Solution
A method utilizing a movable lens and image sensor to acquire multiple images with different blur quantities, simulating gaussian blur and relating it to pillbox blur, allowing for the calculation of a non-linear blur difference to determine depth maps, applicable in various imaging devices like cameras and smartphones.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional hill-climbing or focal gradient methods are used for depth map determination, then autofocusing capability is achieved, but the methods require controlled lighting conditions and high light intensity discrimination
Solution Approach 1:
The patent replaces optical-based autofocus methods (hill-climbing, focal gradient) that depend on light intensity with a depth-from-defocus computational method. By capturing images at multiple focal planes and using computational algorithms to analyze blur characteristics, the system determines depth information without relying on controlled lighting conditions or high light intensity discrimination, thus resolving the contradiction between reliability and adaptability to lighting conditions
Solution Approach 2:
The patent transitions from two-dimensional image analysis (single focal plane) to three-dimensional depth map generation by capturing images at multiple focal planes. This dimensional extension allows the system to extract depth information through computational analysis of focus variation across planes, eliminating dependence on lighting conditions that constrain traditional 2D autofocus methods
2Measurement precision
If Fourier transform methods are used to compute depth from multiple images, then depth information is obtained, but the computational cost is very expensive involving complex and costly hardware
Solution Approach 1:
The patent extracts only the essential depth information from multi-plane images using simplified computational algorithms rather than performing complete Fourier transforms. By focusing on blur characteristic analysis and depth-from-defocus calculations, the system obtains accurate depth measurements while avoiding the computational overhead and hardware complexity of full Fourier transform implementations
Solution Approach 2:
The patent employs computationally efficient algorithms that can be implemented in standard processors rather than requiring specialized, expensive Fourier transform hardware. The method uses accessible computational resources to achieve depth mapping, replacing complex hardware requirements with software-based solutions that are both accurate and cost-effective
3Measurement precision
If Pentland's edge-based depth-map recovery method is used, then depth estimation is achieved, but the method cannot be used if there are no perfect step edges in the scene
Solution Approach 1:
The patent replaces edge-based depth estimation (which requires perfect step edges) with a depth-from-defocus method that analyzes blur characteristics across multiple focal planes. This substitution allows depth estimation to work with natural, continuous scene variations rather than requiring discrete edge features, significantly improving adaptability to different scene types while maintaining depth estimation accuracy
Solution Approach 2:
The patent changes the fundamental parameter used for depth estimation from edge location (Pentland's method) to blur magnitude and focus variation across multiple planes. By measuring how blur characteristics change with focal plane position, the system can estimate depth for any scene content regardless of edge presence, resolving the contradiction between measurement precision and scene feature requirements
Data Source
AI summary
A method of and an apparatus for determining a depth map utilizing a movable lens and an image sensor are described herein. The depth information is acquired by moving the lens a short distance and acquiring multiple images with different blur quantities. An improved method of simulating gaussian blur and an approximation equation that relates a known gaussian blur quantity to a known pillbox quantity are used in conjunction with a non-linear blur difference equation. The blur quantity difference is able to be calculated and then used to determine the depth map. Many applications are possible using the method and system described herein, such as autofocusing, surveillance, robot/computer vision, autonomous vehicle navigation, multi-dimensional imaging and data compression.