Blur Detection in Digital Images via Edge Convolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting and quantifying blur in digital images are either too slow or not reliable, particularly for 'analog' blur, and often rely on arbitrary brightness thresholds, making them inefficient for large images with billions of pixels.
Innovation Solution
A method using a computer program that calculates two scores, S1 and S2, through convolution operations with an edge detection matrix and a logistic function, to determine the blur in digital images, eliminating the need for brightness thresholds and improving processing speed and reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If wavelet transform or Fourier transform methods are used for blur detection, then measurement precision is improved, but productivity deteriorates due to computation time of 500 ms to 3 s per image
Solution Approach 1:
The patent extracts only the essential information needed for blur detection by using simple convolution operations with edge detection matrices, rather than applying comprehensive transform methods. This extraction approach retrieves the critical blur-related features while discarding unnecessary computational steps, achieving fast processing without sacrificing detection accuracy
Solution Approach 2:
The patent replaces complex mathematical transform systems (wavelet/Fourier transforms) with simpler convolution-based edge detection mechanisms. This substitution uses basic linear algebra operations that are computationally lighter, maintaining the ability to detect blur while dramatically reducing computation time to enable real-time processing
2Productivity
If simple methods like color intensity analysis are used, then productivity is improved with processing time of about 100 ms per image, but measurement precision deteriorates due to low reliability
Solution Approach 1:
The patent introduces an intermediary edge detection matrix as a mediator between the simple convolution operation and the final blur score. This matrix acts as a filter that enhances edge information and suppresses irrelevant details, enabling reliable blur detection even with the simplified computational approach
Solution Approach 2:
The patent changes the parameter being measured from general color intensity to edge-specific brightness parameters obtained through convolution with edge detection matrices. This parameter transformation allows the simple method to capture the essential characteristics of blur (edge degradation) while maintaining fast processing speed
3Device complexity
If brightness threshold processing is used for blur quantification, then device complexity is reduced, but measurement precision deteriorates due to arbitrary threshold definition affecting objectivity
Solution Approach 1:
Instead of setting arbitrary brightness thresholds and checking which pixels exceed them, the patent inverts the approach by convolving the image with edge detection matrices to obtain edge-specific brightness parameters. This inversion eliminates the need for threshold selection while maintaining objectivity, as the convolution operation naturally emphasizes edge regions without requiring manual threshold calibration
Data Source
AI summary
A method of detecting and quantifying blur in a digital image, making use of a computer and comprising:a step a) of obtaining a digital image comprising pixels;a step b) of obtaining a brightness parameter for each pixel, said step b) comprising a convolution operation with an edge detection matrix;a step c) of calculating a score S1 comprising the maximum, calculated over all the pixels, of the brightness parameter obtained in step b) and a second score S2 comprising a logistic function, anda step d) of evaluating the digital image, said digital image being considered to be blurry if the score S1 obtained in step c) is strictly less than a first predetermined threshold S10, the score S1 providing a first quantity of blur present in the digital image.