Blur Image Adjusting Method Using Edge Detection and PSF Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current anti-vibration methods for digital cameras, such as mechanical structures and high ISO sensitivity, are costly and complex, making them unsuitable for handheld devices like mobile phones or PDAs, and existing image restoration algorithms like the Lucy-Richardson algorithm are computationally intensive, while direct inverse filters are prone to noise amplification.
Innovation Solution
A blur image adjusting method that converts images to the YCbCr color space, extracts the luminance component, uses edge detection to estimate horizontal and vertical shift amounts, determines a point spread function, and applies a restoration filter like the Wiener filter to recover sharp images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If mechanical anti-vibration structures are used, then image sharpness is improved, but device complexity and cost increase
Solution Approach 1:
The patent replaces mechanical anti-vibration structures with a digital image processing approach. Instead of using physical mechanisms to prevent vibration during capture, the system captures blurred images and applies computational algorithms (point spread function analysis and restoration filtering) to recover sharp images digitally, thereby eliminating complex mechanical components.
Solution Approach 2:
The patent creates a digital model (point spread function) that represents the vibration blur characteristics. By analyzing the blur pattern and creating a corresponding PSF model, the system can digitally replicate and reverse the degradation process through restoration filtering, avoiding the need for physical anti-vibration mechanisms.
2Manufacturing precision
If high ISO sensitivity is used for anti-vibration, then image sharpness is improved, but noise amplification occurs
Solution Approach 1:
The patent replaces the high ISO sensitivity approach (which amplifies signals including noise) with a digital restoration approach. Instead of amplifying weak signals during capture, the system captures the full-dynamic-range image and applies restoration filtering to sharpen the image afterward, avoiding noise amplification while achieving the same sharpness goal.
Solution Approach 2:
The patent performs the anti-vibration function in advance by capturing the complete image data without interruption, then applies restoration filtering as a preliminary processing step before final image output. This allows the system to address vibration blur without the need for real-time signal amplification during capture.
3Manufacturing precision
If direct inverse filter is used for image restoration, then image sharpness is improved, but noise is amplified
Solution Approach 1:
The patent modifies the restoration filter parameters by introducing a regularization term (epsilon) to the inverse filter equation. This parameter change transforms the unstable direct inverse filter into a stable restoration filter that suppresses noise amplification while maintaining image sharpness, effectively balancing restoration quality and noise control.
4Manufacturing precision
If complex restoration algorithms like Lucy-Richardson are used, then image sharpness is improved, but processing time increases
Solution Approach 1:
The patent extracts only the essential characteristics of the blur (represented by the point spread function) and applies a targeted restoration filter based on these extracted features. Instead of using computationally intensive iterative algorithms, the system extracts the dominant blur pattern and applies a efficient restoration filter that achieves good results with minimal processing time.
Data Source
AI summary
A blur image adjusting method includes the following steps. Firstly, a blur image in YCbCr color space is obtained. The Y component of the blur image is extracted so as to obtain a Y component blur image. A blur area is extracted from the Y component blur image by an edge detection technology. A horizontal shift amount and a vertical shift amount are estimated according to a horizontal shift pixel number distribution and a vertical shift pixel number distribution of the blur area. A point spread function is determined according to the horizontal shift amount and the vertical shift amount. Afterwards, the blur image is adjusted according to the point spread function.