Deconvolution Filter for Moving Object Image De-blurring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automatic recognition systems for objects like license plates suffer from blurring when the object is moving, leading to degraded optical character recognition due to image displacement during camera exposure, especially when the object moves by at least a pixel during image capture.
Innovation Solution
A method involving the capture of two successive images by a digital video camera, where a motion vector is determined to assess blurring, and a deconvolution filter is applied using a point spread function based on the exposure time and frame interval to de-blur the object region, allowing for improved optical character recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a digital camera captures an image of a moving object during exposure, then the image is obtained, but the object becomes blurred due to movement
Solution Approach 1:
The system performs preliminary actions by capturing a sequence of images before the final image is needed, tracking object motion in advance. Motion vectors are calculated from previous frames to predict and compensate for blur in the current frame, allowing the system to prepare deconvolution filters before the blurred image would otherwise be unusable.
Solution Approach 2:
The patent introduces intermediary elements including motion vectors as mediators that transfer motion information between frames, and deconvolution filters as mediators that transfer the blurred image to a de-blurred version. These intermediaries enable the system to handle moving objects by processing motion information separately and applying it to restore image quality.
2Illumination intensity
If the shutter is opened for a longer duration to capture more light, then image brightness is improved, but motion blur increases
Solution Approach 1:
The system converts the harmful effect of motion blur (caused by longer exposure) into a beneficial signal. By detecting and quantifying the blur through motion vector analysis, the system creates an accurate model of the degradation, which then enables precise mathematical reversal through deconvolution, ultimately improving both brightness and sharpness simultaneously.
Solution Approach 2:
The patent changes parameters by dynamically adjusting deconvolution filter coefficients based on calculated motion vectors. Instead of using a fixed processing approach, the system adapts processing parameters (filter strength, kernel size) to match the actual motion magnitude and direction, allowing optimal recovery of image quality across varying exposure conditions.
3Manufacturing precision
If de-blurring processing is applied to remove motion blur, then image clarity is improved, but processing complexity increases
Solution Approach 1:
The system applies partial action by focusing deconvolution processing only on regions containing moving objects rather than processing the entire image. By identifying object boundaries and limiting processing to relevant areas, the system achieves effective de-blurring while significantly reducing computational complexity and processing time.
Solution Approach 2:
The patent segments the image processing task into distinct components: motion detection, motion vector calculation, deconvolution filter application, and result combination. This segmentation allows each component to be optimized independently and enables parallel processing of different image regions, reducing overall system complexity while maintaining high processing quality.
Data Source
AI summary
A method and apparatus for object-of-interest image de-blurring includes functions that determine (305, 310) an object region of an object-of-interest in an image that may be generated by image capture device (805), that determine (315, 825) a motion vector of the object-of-interest, that determine (325, 830) a scaling factor, and that generate (330) a de-blurred object region by processing the object region using a deconvolution filter (835) formulated from the motion vector and the scaling factor.