Estimation of motions in color image sequences using hypercomplex fourier transforms

Area of Science:

• Image Processing
• Computer Vision
• Signal Processing

Background:

• Motion estimation is crucial for video analysis but often limited to monochrome sequences.
• Current color video methods typically process channels independently, which is suboptimal.
• Quaternion-based approaches offer holistic color processing but can be complex.

Purpose of the Study:

• To extend spatiotemporal Fourier-based methods for holistic color image sequence motion estimation.
• To develop a methodology for motion estimation using the hypercomplex Fourier transform.
• To compare the efficiency and effectiveness against channel-separate processing.

Main Methods:

• Utilized the hypercomplex Fourier transform to analyze color image sequences holistically.
• Identified translational motion signatures as energy concentrations in the 3-D hypercomplex Fourier domain.
• Developed a motion estimation technique based on these identified properties.

Main Results:

• Demonstrated that translational motions manifest as planar energy concentrations in the hypercomplex Fourier domain.
• Achieved a computational effort reduction of one-third compared to processing channels separately.
• Proposed an effective method for extracting moving objects within the hypercomplex Fourier domain.

Conclusions:

• The hypercomplex Fourier transform provides an efficient and effective framework for color video motion estimation.
• This holistic approach surpasses traditional channel-separate processing in terms of computational efficiency.
• The method facilitates robust motion estimation and moving object extraction in color image sequences.