Graph diffusion based on Quantum cuts algorithm for butterfly segmentation

Published: 25 August 2022| Version 1 | DOI: 10.17632/jvgvp3my5g.1
Idir Filali


We present the contrast maps results obtained by our graph based diffusion process that exploits efficiently wave functions in the Schrodinger equation in a mono-layered undirected graph using the public Leeds Butterfly dataset (J. Wang, K. Markert, M. Everingham, Learning models for object recognition from natural language descriptions. Proc. br. Mach. Vis. Conf., (2009), pp. 2.1-2.11, The dataset contains 832 ecological images on which each one represents a single butterfly in its natural environment. There is ten species of butterflies in the dataset and a distribution of 55 to 100 photographs is repertoried in each of them. The images are characterized by several artifacts that complicate the segmentation task like the presence of irrelevant objects, the shadow effect, the variation of illumination and the high similarity between some background and butterfly regions. Our algorithm evaluates the regional contrast based on Quantum cuts theory by performing a diffusion process of intensity scores from the foreground template towards the image borders. The obtained contrast maps allow to extract efficiently the butterfly regions and supress the background areas. In order to evaluate visually the performance of our algorithm, ground truth images are provided. Folders descriptions: -> Images: Color ecological images. -> Ground Truth: The real position of butterflies in the images. -> Contrast Maps: Results returned by our algorithm. -> Contrast Estimation: a compressed folder containing the above folders



Universite Mouloud Mammeri de Tizi Ouzou


Image Segmentation, Ecological Engineering, Precision Agriculture, Computer Vision Algorithms