Entropy thresholding and its parallel algorithm on the reconfigurable array of processors with wider bus networks

Area of Science:

• Computer Vision
• Image Processing
• Parallel Computing

Background:

• Image segmentation is crucial in image processing, with thresholding being a widely adopted technique.
• Existing relative entropy-based thresholding methods can be computationally intensive.
• Parallel processing architectures offer potential for accelerating image segmentation algorithms.

Purpose of the Study:

• To develop an efficient sequential algorithm for improving relative entropy-based thresholding.
• To design a constant-time parallel algorithm for entropy-based thresholding.
• To leverage reconfigurable array of processors with wider bus networks (RAPWBN) for enhanced performance.

Main Methods:

• An efficient sequential algorithm combining relative entropy, local entropy, and quadtree hierarchical structure was proposed.
• A constant-time parallel algorithm was derived for the RAPWBN architecture.
• The RAPWBN architecture utilizes wider bus networks to enhance data communication and processing power, optimizing silicon area usage.

Main Results:

• The proposed sequential algorithm improves upon existing relative entropy-based thresholding techniques.
• The parallel algorithm achieves constant-time complexity on the RAPWBN.
• The RAPWBN architecture demonstrates significant advantages in silicon area efficiency and system power compared to traditional parallel systems.

Conclusions:

• The developed algorithms offer efficient solutions for entropy-based image thresholding.
• The RAPWBN architecture is a powerful and area-efficient platform for parallel image processing tasks.
• This research contributes to advancements in image segmentation and parallel computing strategies.