New multispectral sensors, which are capable of capturing high resolution images through low altitude drone flights, offer access to a wealth of information about the Earth's surface at a significantly lower cost than other imaging devices. The process of identifying unexpected patterns within an image that do not conform to the expected behavior is known as anomaly detection (AD). When applied to fluvial ecosystem monitoring, this involves detecting the existence of small constructions or roads that allow automatic alarms to be produced for the people in charge of monitoring the ecosystem. The extraction of spatial information is a critical step in AD, since it determines the final quality of the AD and it is a computationally expensive processing. In this work, Extinction Profiles (EP) are selected to perform a multilevel implicit segmentation of the image, thus extracting the spatial information of relevance. A computationally efficient implementation of the EP-based spatial extraction of information for multidimensional images is proposed in this paper, as it is a basic step in the detection of anomalies in natural ecosystems. The proposed method takes advantage of heterogeneous computing to perform the task in a reduced execution time.

This work was supported in part by grants PID2019--104834GB--I00, PID2022-141623NB-I00, and TED2021--130367B--I00 funded by MCIN/AEI/10.13039/501100011033 and by European Union NextGenerationEU/PRTR. It was also supported by Xunta de Galicia - Consellería de Cultura, Educación, Formación Profesional e Universidades [Centro de investigación de Galicia accreditation 2019-2022 ED431G-2019/04 and Reference Competitive Group accreditation, ED431C-2022/16], by Junta de Castilla y León [Project VA226P20 (PROPHET--II)], and by European Regional Development Fund (ERDF).