Parallel segmentation algorithm over aquaculture images

Abstract

Within the field of aquaculture, when working with image analysis, the correct execution of preprocessing and segmentation algorithms for the selection of fish regions becomes a primary factor. Depending on the case, the number of images to be processed by these algorithms can significantly increase the time required to obtain the final result, thus delaying the ability to provide results. When it comes to processing large amounts of data, the paradigm of parallelization has given important results, reducing the execution time to complete complex tasks. In this article, it is proposed to carry out the preprocessing and segmentation processes in parallel, trying to optimize the time required to obtain the set of images prepared for their subsequent analysis. A comparison was made in execution times with different computers, and the experimental results show that it is better to
use four or six execution threads by parallelizing the image preprocessing and segmentation algorithm.