Foveal scale space generation with the log-polar transform

Tracking deep-field objects across a wide field of view requires the use of high resolution image sensors. This imposes a burden on processing systems which must detect and extract features in an image. Deep-field objects have limited spatial support within a wide field of view image and accordingly much of the recorded scene contains superfluous information about the environment. This paper explores the generation of a foveal scale space using the log-polar transform. Foveal scale space is the scale space representation of an input scene where the spatial support of the image at each scale increases with scale and the number of pixels remains constant across each slice of scale space. This paper reports the formulation of a transformation consisting of a peripheral region defined by the log-polar transform and a foveal region where resolution is constant. A method for performing diffusion in this domain is shown and the generation of the foveal scale space is presented.

[1]  Andrew P. Witkin,et al.  Scale-space filtering: A new approach to multi-scale description , 1984, ICASSP.

[2]  Martin D. Levine,et al.  A Review of Biologically Motivated Space-Variant Data Reduction Models for Robotic Vision , 1998, Comput. Vis. Image Underst..

[3]  H. Araujo,et al.  An introduction to the log-polar mapping [image sampling] , 1996, Proceedings II Workshop on Cybernetic Vision.

[4]  J. Koenderink.  The structure of images , 2004, Biological Cybernetics.

[5]  G. Griffin,et al.  Caltech-256 Object Category Dataset , 2007 .

[6]  Alexandre Bernardino,et al.  A review of log-polar imaging for visual perception in robotics , 2010, Robotics and Autonomous Systems.

[7]  T. Lindeberg.  Scale-Space Theory : A Basic Tool for Analysing Structures at Different Scales , 1994 .

[8]  R. Matungka,et al.  2D invariant object recognition using Log-Polar transform , 2008, 2008 7th World Congress on Intelligent Control and Automation.

[9]  Alexandre Bernardino,et al.  Smooth Foveal vision with Gaussian receptive fields , 2009, 2009 9th IEEE-RAS International Conference on Humanoid Robots.