Blur Direction Judgment Using Active Appearance Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting camera shake in images, such as those using angular rate sensors, are not feasible for small devices like camera phones and do not provide a solution for detecting shake based on image data alone.
Innovation Solution
A method that uses statistical characteristic quantities and weighting parameters to determine the direction of blur in an image by fitting a model to the image data, without requiring special hardware like angular rate sensors, utilizing techniques like Active Appearance Models for image processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If an angular rate sensor is installed to detect camera shake, then detection accuracy is improved, but device complexity and size increase
Solution Approach 1:
The patent replaces the mechanical/angular rate sensor system with a computational image processing system. By analyzing blur characteristics in the captured image itself, the system determines camera shake direction without requiring any additional sensing hardware. This substitution of mechanical detection with optical-computational analysis resolves the contradiction between detection accuracy and device complexity.
Solution Approach 2:
The patent creates a computational model of blur patterns that represents the physical camera shake phenomenon. By fitting statistical models to image data and extracting weighting parameters that represent blur characteristics, the system creates a virtual representation of camera shake that can be analyzed and corrected without physical sensors.
2Reliability
If an angular rate sensor is installed to detect camera shake, then detection reliability is improved, but ease of manufacture deteriorates
Solution Approach 1:
The patent eliminates the need for installing and calibrating angular rate sensors by using standard image processing pipelines already present in digital cameras. The method uses conventional image data and statistical analysis to achieve reliable camera shake detection, making the system much easier to manufacture and deploy in standard digital cameras without special hardware modifications.
3Ease of manufacture
If image processing methods are used to detect camera shake, then ease of manufacture is improved, but measurement precision deteriorates
Solution Approach 1:
The patent employs sophisticated statistical parameter extraction by fitting models to image data and analyzing weighting parameters that represent blur characteristics. By changing from simple image analysis to advanced statistical parameter estimation, the method achieves high measurement precision while maintaining ease of manufacture through software-based processing.
Solution Approach 2:
The patent applies different processing strategies to different regions and characteristics of the image. By analyzing local blur patterns and their statistical properties in specific image regions, the method extracts precise camera shake information while using standard image processing techniques that are easy to implement.
Data Source
AI summary
Without use of special hardware such as an angular rate sensor, a direction of blur can be judged with high accuracy. A parameter acquisition unit obtains weighting parameters for principal components representing directions of blur in a predetermined structure in an image by fitting to the structure a mathematical model generated by a method of Active Appearance Models (AAM) using a plurality of sample images representing the structure in different conditions of blur. A blur direction judgment unit judges the direction of blur based on the weighting parameters, and a blur width acquisition unit finds a width of blur based on an edge component found by an edge detection unit in the direction perpendicular to the direction of blur. A blur correction unit corrects the blur in the image based on the direction and the width of blur.