In this paper we present a novel filter, based on the existing COSFIRE filter, for the delineation of patterns of interest. It includes a mechanism of push-pull inhibition that improves robustness to noise in terms of spurious texture. Push-pull inhibition is a phenomenon that is observed in neurons in area V1 of the visual cortex, which suppresses the response of certain simple cells for stimuli of preferred orientation but of non-preferred contrast. This type of inhibition allows for sharper detection of the patterns of interest and improves the quality of delineation especially in images with spurious texture. We performed experiments on images from different applications, namely the detection of rose stems for automatic gardening, the delineation of cracks in pavements and road surfaces, and the segmentation of blood vessels in retinal images. Push-pull inhibition helped to improve results considerably in all applications.
CITATION STYLE
Strisciuglio, N., Azzopardi, G., & Petkov, N. (2019). Brain-inspired robust delineation operator. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 11131 LNCS, pp. 555–565). Springer Verlag. https://doi.org/10.1007/978-3-030-11015-4_41
Mendeley helps you to discover research relevant for your work.