Dual tree fractional quaternion wavelet transform for disparity estimation.

Area of Science:

• Computer Vision
• Image Processing
• Signal Analysis

Background:

• Optical flow and disparity estimation are crucial for understanding motion and depth in image sequences.
• Existing methods often face challenges with computational efficiency and accuracy, especially in handling complex image features.

Purpose of the Study:

• To propose a novel phase-based approach for computing disparity and optical flow.
• To introduce a new dual tree fractional quaternion wavelet transform (FrQWT) for enhanced image analysis.

Main Methods:

• Development of a new dual tree fractional quaternion wavelet transform (FrQWT) by defining the 2D Fourier spectrum in a single quadrant.
• Utilizing the real and imaginary parts of the FrQWT to encode horizontal, vertical shifts, and texture information.
• Implementing a multi-scale framework for disparity calculation and phase unwrapping from coarse to fine scales.

Main Results:

• The proposed FrQWT effectively encodes image feature shifts and texture information.
• The method allows for multi-scale adjustment of local disparities.
• Phase unwrapping is executed efficiently from coarse to fine scales with linear computational complexity.

Conclusions:

• The novel phase-based FrQWT approach offers an efficient and accurate method for disparity and optical flow computation.
• This technique provides a robust framework for multi-scale image analysis and feature extraction.