Correction of geometrically distorted underwater images using shift map analysis

Area of Science:

• Computer Vision
• Image Processing
• Optical Engineering

Background:

• Underwater images suffer from geometric distortions and blurring due to water waves.
• Existing methods for distortion correction require improved pixel shift map estimation for higher accuracy.

Purpose of the Study:

• To develop an efficient algorithm for estimating pixel shift maps from distorted underwater video sequences.
• To restore geometrically distorted and blurred underwater images with enhanced accuracy.

Main Methods:

• A nonrigid image registration method estimates pixel shift maps between distorted frames and a reference frame.
• The sharpest frame, identified by a sharpness metric, serves as the reference.
• K-means clustering is used to exclude blurry frames, ensuring accurate shift map estimation.

Main Results:

• The algorithm efficiently estimates accurate pixel shift maps for underwater video sequences.
• Restored sequences show significant improvements in clarity and geometric accuracy compared to state-of-the-art methods.
• The method demonstrates effectiveness on both synthetic and real-world underwater video data.

Conclusions:

• The proposed algorithm offers a robust solution for correcting water wave-induced distortions in underwater imaging.
• Accurate pixel shift map estimation is crucial for high-fidelity underwater image restoration.
• This work advances the state-of-the-art in underwater image dewarping and restoration.