Classifying Discriminative Features for Blur Detection.

Area of Science:

• Computer Vision
• Image Processing
• Machine Learning

Background:

• Blur detection in single images, particularly with spatially-varying blur, remains a significant challenge.
• Developing effective and discriminative blur features is an ongoing research problem in image analysis.

Purpose of the Study:

• To propose a novel kernel-specific feature vector for improved blur detection.
• To address the limitations of existing methods in handling complex blur types.

Main Methods:

• A new kernel-specific feature vector is proposed, combining blur kernel and image patch information.
• The feature is derived from the multiplication of the variance of filtered kernel and filtered patch gradients, grounded in a blur-classification theorem.
• A comprehensive pool of motion-blur, defocus-blur, and combined kernels was created for training and testing.

Main Results:

• The proposed kernel-specific features demonstrate strong discriminative power for blur classification.
• Algorithms utilizing these features significantly outperform state-of-the-art blur detection methods.
• Experimental validation on public databases confirms the method's effectiveness and robustness.

Conclusions:

• The proposed kernel-specific feature vector offers a robust and effective solution for single-image blur detection.
• This approach advances the field by providing a theoretically grounded and practically effective feature for analyzing spatially-varying blur.
• The method shows high potential for real-world applications requiring accurate blur identification.