Time Optimization Strategies for Image Segmentation on Modern Multicore Architectures

Recent advancements in computer engineering have led to the development of multicore processors—chips that integrate two or more independent processing units (cores) within a single physical package. With growing demand for high-performance computing, multicore designs have become standard in modern processors. These architectures offer significant opportunities for performance gains, especially in fields like image processing. By adapting traditionally single-threaded code to run in parallel, developers can fully leverage the capabilities of multicore systems. However, maximizing the performance of multicore architectures poses challenges. Beyond achieving raw speed, it is crucial to manage and utilize computing resources efficiently. Unlike older, single-core systems where performance improvements could be expected through hardware upgrades alone, modern applications must be redesigned using parallel algorithms to truly benefit from multicore systems. Image processing, which naturally involves tasks that can be executed concurrently, is particularly well-suited to parallelization. This dissertation focuses on harnessing the power of multicore systems for image processing tasks, with the primary goal of minimizing execution and wait times. To achieve this, we employed the fork-join model of parallel computing, enabling jobs to be executed simultaneously across multiple cores. As part of this research, image segmentation algorithms were implemented using MATLAB. The study aims to improve overall system efficiency by reducing execution time and increasing responsiveness. Ultimately, the work demonstrates how parallel algorithm design can unlock the full potential of multicore processors in image processing applications.