Sharpness Classifier Using Frequency Band Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image evaluation systems fail to accurately and consistently assess the sharpness of images, particularly for challenging types such as high noise, macros, close-up portraits, and night scenes, due to movement and focus issues during exposure.
Innovation Solution
A method involving a sharpness classifier trained on sharpness features computed from preprocessed images, using high pass and band pass filtering, texture region identification, and metadata analysis to determine if an image is sharp or blurred, with adjustments for specific image types like motion blur, high noise, and low exposure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing sharpness evaluation methods are used, then the evaluation process is simple, but the accuracy and consistency of sharpness assessment deteriorates, especially for challenging image types
Solution Approach 1:
The image is divided into multiple frequency bands using high-pass and band-pass filtering. The evaluation system segments the image into different frequency components (high frequency for fine details, band frequency for intermediate structures) and evaluates sharpness separately in each band, then combines the results. This segmentation allows accurate assessment of different image characteristics while managing computational complexity through focused analysis of specific frequency ranges.
Solution Approach 2:
The patent transforms the sharpness evaluation from a single-dimensional metric to a multi-dimensional assessment by introducing frequency domain analysis. Instead of evaluating sharpness directly in spatial domain, the system converts images to frequency domain using filters, creating additional evaluation dimensions (frequency bands, texture regions, energy distributions) that provide more comprehensive and accurate sharpness measurement.
2Adaptability or versatility
If conventional sharpness metrics are applied to all images, then the processing is fast, but the performance deteriorates for specific image types such as high noise, macros, close-up portraits and night scenes
Solution Approach 1:
The evaluation system applies different processing strategies to different regions and image types. Texture regions are identified and evaluated separately from non-texture regions. Different frequency bands are weighted differently based on their contribution to perceived sharpness. The system adapts its evaluation criteria locally according to image characteristics, providing accurate assessment for diverse image types including high noise, macros, portraits, and night scenes.
Solution Approach 2:
The patent implements a dynamic evaluation system that adapts its parameters and methods based on the input image characteristics. The system dynamically adjusts filtering parameters, region-of-interest selection, and evaluation weights according to the specific image type and content. This dynamic adaptation enables the system to maintain high accuracy across varying image conditions while optimizing processing efficiency for each case.
3Measurement precision
If multiple processing steps are applied to improve accuracy, then the sharpness detection precision improves, but the computational complexity increases
Solution Approach 1:
The system performs preliminary processing steps including high-pass and band-pass filtering, texture region identification, and energy calculation before the final sharpness evaluation. By preparing frequency-separated images and identifying relevant texture regions in advance, the system reduces the complexity of the final decision-making process and enables more accurate measurement through pre-organized data structures and pre-computed features.
Data Source
AI summary
A method for predicting whether a test image (318) is sharp or blurred includes the steps of: providing a sharpness classifier (316) that is trained to discriminate between sharp and blurred images; computing a set of sharpness features (322) for the test image (318) by (i) generating a high pass image (404) from the test image (318), (ii) generating a band pass image (406) from the test image (318), (iii) identifying textured regions (408) in the high pass image, (iv) identifying texture regions (410) in the band pass image, and (v) evaluating the identified textured regions in the high pass image and the band pass image to compute the set of test sharpness features (412); and evaluating the sharpness features using the sharpness classifier (324) to estimate if the test image (318) is sharp or blurry (20).