Fast Graph Partitioning Active Contours for Image Segmentation Using Histograms

Area of Science:

• Computer Vision
• Image Processing
• Computational Mathematics

Background:

• The Graph Partitioning Active Contours (GPACs) algorithm offers a novel approach to image segmentation.
• Existing GPACs implementations face challenges with computational expense and memory requirements, particularly for large images.
• Improvements are needed to enhance the efficiency and applicability of GPACs.

Purpose of the Study:

• To develop a computationally efficient reformulation of the GPACs algorithm.
• To significantly reduce the memory footprint of the GPACs algorithm.
• To improve the accuracy and speed of image segmentation using GPACs.

Main Methods:

• Replaced the computation of an approximate dissimilarity matrix with fixed length histograms and a symmetric-centrosymmetric extensor matrix.
• Jointly computed terms associated with the complete dissimilarity matrix.
• Developed a computationally efficient reformulation of GPACs with a reduced memory footprint.

Main Results:

• Achieved significant reductions in memory requirements compared to the original GPACs algorithm.
• Demonstrated accelerated convergence of the evolving active contour.
• Extended the performance of GPACs to multidimensional images seamlessly.

Conclusions:

• The reformulated GPACs algorithm offers substantial improvements in efficiency and performance.
• This method provides a more practical and scalable solution for image segmentation tasks.
• The enhanced GPACs algorithm is suitable for both 2D and multidimensional image analysis.