Work

TurboPixels: A Superpixel Segmentation Algorithm Suitable for Real-Time Embedded Applications

Year: 2024

Type: article

Abstract: Superpixel segmentation aims to produce a consistent grouping of pixels. In recent years, the importance of superpixel segmentation has increased in computer vision since it offers useful primitives for extracting image features and simplifies the complexity of other image processing steps. In this work, we propose the TurboPixels algorithm, whose main contribution is a hardware architecture for superpixel segmentation. Compared with previous approaches, our superpixels are computed without the need for iterative loops. This makes it possible to reduce algorithmic complexity and increases processing speed. The experimental results indicate that our approach enables a small-scale FPGA-based implementation suitable for embedded applications. In addition, the results demonstrate that robust superpixel segmentation can be achieved with processing speeds up to 86 times faster than in previous works in the current literature.

Source: Applied Sciences

DOI: 10.3390/app142411912

Authors: Abiel Aguilar-González, Alejandro Medina Santiago, Jorge Antonio Orozco Torres, J. A. de Jesús Osuna-Coutiño, Madaín Pérez Patricio, Néstor A. Morales-Navarro

Institutions: National Institute of Astrophysics, Optics and Electronics, Tuxtla Gutierrez Institute of Technology, Tecnológico Nacional de México

Language: English

Topic: Medical Image Segmentation Techniques

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

SDG: Industry, innovation and infrastructure

Open Access status: gold

Estimated APC paid: $2,728