Graph theory approach for match reduction in image mosaicing

Area of Science:

• Computer Vision
• Robotics
• Computational Geometry

Background:

• Global alignment is critical for image mosaicing, often requiring computationally intensive nonlinear minimization over image pair correspondences.
• Existing methods for image registration can be resource-demanding, especially for large image datasets.

Purpose of the Study:

• To propose a novel, efficient method for reducing the number of overlapping image pairs in image mosaicing.
• To decrease the computational cost of the global alignment process in image mosaicing.
• To maintain the final mosaic quality while optimizing the registration process.

Main Methods:

• Utilizing graph theory to identify and reduce redundant overlapping image pairs within a dataset.
• Implementing a topology estimation process to minimize image matching attempts.
• Applying nonlinear minimization methods for image registration parameter optimization.

Main Results:

• Successfully reduced the number of overlapping image pairs significantly.
• Demonstrated no noticeable degradation in the final mosaic quality.
• Achieved a substantial reduction in the overall computational cost of image mosaicing.

Conclusions:

• The proposed graph theory-based method offers an efficient solution for optimizing image mosaicing.
• The technique is effective in reducing computational load without compromising mosaic quality.
• Validated on challenging underwater image sequences, showing broad applicability.