Temporal Coherence-Based Deblurring Using Non-Uniform Motion Optimization.

Area of Science:

• Computer Vision
• Image Processing
• Video Analysis

Background:

• Non-uniform motion blur is a prevalent issue in videos, often caused by object or camera movement.
• Current video deblurring methods struggle with complex motion blur, sometimes producing artifacts or failing entirely.
• Existing techniques may not adequately address rotational, scale, or blur variations.

Purpose of the Study:

• To propose a novel non-uniform motion model for video deblurring.
• To enhance the reconstruction of sharp video frames from blurry inputs.
• To overcome limitations of current state-of-the-art video deblurring methods.

Main Methods:

• Utilizing superpixel matching within video sequences for frame reconstruction.
• Employing a generalized PatchMatch algorithm with non-uniform motion blur kernels to handle diverse blur types.
• Adopting a superpixel-based representation, leveraging color and motion for pixel similarity.
• Estimating non-uniform motion blur kernels from superpixel motion fields and incorporating spatial-temporal coherence.

Main Results:

• The proposed method successfully reconstructs sharp video frames from inputs with complex object and camera movements.
• Demonstrated superior performance compared to existing state-of-the-art video deblurring techniques.
• Effectively handles rotation, scale, and blur differences during the matching process.

Conclusions:

• The developed non-uniform motion model offers a robust solution for video deblurring.
• Superpixel matching provides an effective representation for handling complex motion blur.
• The method shows significant improvements in deblurring performance for challenging video sequences.