Design and implementation of the Hough transform on GPU using CUDA

Abstract

[Abstract]: The Hough transform is a robust algorithm that is used to detect straight lines in binary images, even in the presence of noise and occlusion. However, it is computationally expensive. This project proposes the implementation of this algorithm on a GPU (Graphics Processing Unit), taking advantage of the massive parallel architecture to achieve significant performance over the CPU (Central Processing Unit). Various optimization techniques and thread configurations for processing multiple images simultaneously will be explored to evaluate the effectiveness of the proposed implementations and identify potential areas for further improvement.

[Resumo]: A transformada de Hough é un algoritmo robusto que se utiliza para detectar liñas rectas en imaxes binarias, mesmo na presenza de ruído e oclusión. Non obstante, ten un alto custo computacional. Este proxecto propón a implementación desde algoritmo nunha GPU, aproveitando a arquitectura paralela masiva para acadar unha mellora significativa no rendemento en comparación coa CPU. Exploraranse diversas técnicas de optimización e configuracións de fíos para aproveitar os recursos da GPU de forma eficiente.