Frequency Domain Image Processing for Distance Calculation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Depth from Defocus (DFD) techniques face challenges in accurately calculating distance information for fast-moving subjects due to positional deviations between images captured at different focusing positions, leading to difficulties in capturing still or moving images with high accuracy.
Innovation Solution
An image processing device and method that convert images into frequency, extract amplitude components, and use lens blur data to calculate distance information, effectively eliminating the need for precise alignment and enabling accurate distance calculation even with positional deviations, by employing a frequency converter, amplitude extractor, and distance information calculator.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple images with different focusing positions are captured to calculate distance information using conventional DFD, then distance information can be calculated, but positional deviation occurs between images when the subject moves quickly, reducing calculation accuracy
Solution Approach 1:
The patent transforms images from spatial domain to frequency domain using Fourier transform, changing the parameter space from spatial coordinates to frequency components. This parameter transformation allows the system to work with frequency-domain representations where positional deviations become phase shifts that can be handled differently, enabling accurate distance calculation even when subjects move between captures
Solution Approach 2:
The patent extracts only the amplitude component from the frequency-transformed images, separating it from the phase component. By taking out and using only the amplitude information, the system eliminates the problematic phase information that contains positional deviation data, thereby achieving accurate distance measurement without being affected by subject movement between image captures
2Measurement precision
If conventional DFD methods are used to calculate distance information, then distance can be determined from blur information, but the process requires precise alignment of multiple images which is difficult when subjects move
Solution Approach 1:
The patent replaces the mechanical/image alignment process with a frequency-domain processing approach. Instead of physically or computationally aligning images in spatial domain (which requires complex operations when subjects move), the system transforms images to frequency domain where the same alignment function is achieved through phase manipulation, simplifying the overall process while maintaining accuracy for moving subjects
3Loss of information
If multiple images are captured with different focusing positions, then distance information can be calculated, but time difference occurs between captures which causes positional deviation for fast-moving subjects
Solution Approach 1:
By transforming to frequency domain and using amplitude components, the patent changes the temporal sensitivity of the measurement process. The amplitude spectrum in frequency domain represents the magnitude of spatial frequencies independent of temporal position, allowing the system to utilize blur information effectively while becoming insensitive to the time differences between captures of moving subjects
Data Source
AI summary
Image processing device is image processing device that uses a plurality of images respectively having focusing positions different from each other to calculate distance information to a subject, and includes frequency converter, amplitude extractor, and distance information calculator. Frequency converter converts the plurality of images into frequency. Amplitude extractor extracts an amplitude component out of a phase component and the amplitude component of a coefficient obtained by converting the plurality of images into frequency. Distance information calculator calculates the distance information, by using lens blur data and only the amplitude component extracted by amplitude extractor out of the phase component and the amplitude component of the coefficient.