<XML><RECORDS><RECORD><REFERENCE_TYPE>3</REFERENCE_TYPE><REFNUM>8648</REFNUM><AUTHORS><AUTHOR>Balasuriya,L.S.</AUTHOR><AUTHOR>Siebert,J.P.</AUTHOR></AUTHORS><YEAR>2007</YEAR><TITLE>Iconic Object-based Saccade Generation using a Biologically Inspired Self-organized Retina</TITLE><PLACE_PUBLISHED>2007 International Joint Conference on Neural Networks, August 12-17, 2007, Orlando, Florida, USA.</PLACE_PUBLISHED><PUBLISHER>N/A</PUBLISHER><LABEL>Balasuriya:2007:8648</LABEL><KEYWORDS><KEYWORD>space variant vision</KEYWORD></KEYWORDS<ABSTRACT>This paper presents a computer vision attention model which is motivated by many of these processing structures and approaches which have evolved in space-variant biological vision. A novel space-variant pseudo-random artificial retina was used in this work to sample visual information from images. Biologically-motivated space-variant receptive fields extracted visual information for processing in higher stages of our system. Features were extracted as different spatial scales using a pyramid of artificial retinae and associated higher order filters. The extracted features were aggregated into a feature vector that was used as an iconic descriptor for the associated image region in the system’s field-of-view. The vision system targeted the artificial retina on salient locations in the system’s field of view by processing iconic descriptors extracted by the retina. Two methods of attention were used to target the retina : (1) bottom-up attention to help restrict attention of the system to areas in the scene where there was activity among low-level features, and (2) top-down attention which enabled the saccadic fixations of the system to be influenced by the particular task that the system was attempting to perform. Therefore the series of saccadic fixations of the system when it is trying to find a particular object in a scene is quite different to when it is trying to find another object. We will demonstrate the effectiveness of our biologically inspired model using real-world visual scenes and present several scenarios diagnostic of the system's behaviour.</ABSTRACT><NOTES>Invited Paper for workshop on: Biologically-Inspired Computational Vision</NOTES><URL>http://www.ijcnn2007</URL></RECORD></RECORDS></XML>