TY - GEN
T1 - Eccentricity estimator for wide-angle fovea sensor by FMI descriptor approach
AU - Shimizu, Sota
AU - Burdick, J. W.
PY - 2009
Y1 - 2009
N2 - This paper proposes a method for estimating eccentricity that corresponds to an incident angle to a fovea sensor. The proposed method applies Fourier-Mellin Invariant descriptor for estimating rotation, scale, and translation, by taking both geometrical distortion and non-uniform resolution of a space-variant image by the fovea sensor into account. The following 2 points are focused in this paper. One is to use multi-resolution images computed by Discrete Wavelet Transform for reducing noise caused by foveation properly. Another is to use a variable window function (although the window function is generally used for reducing DFT leakage caused by both ends of a signal.) for changing an effective field of view (FOV) in order not to sacrifice high accuracy. The simulation compares the root mean square (RMS) of the foveation noise between uniform and non-uniform resolutions, when a resolution level and a FOV level are changed, respectively. Experimental results show that the proposed method is consistent with the wide-angle space-variant image by the fovea sensor, i.e., it does not sacrifice high accuracy in the central FOV.
AB - This paper proposes a method for estimating eccentricity that corresponds to an incident angle to a fovea sensor. The proposed method applies Fourier-Mellin Invariant descriptor for estimating rotation, scale, and translation, by taking both geometrical distortion and non-uniform resolution of a space-variant image by the fovea sensor into account. The following 2 points are focused in this paper. One is to use multi-resolution images computed by Discrete Wavelet Transform for reducing noise caused by foveation properly. Another is to use a variable window function (although the window function is generally used for reducing DFT leakage caused by both ends of a signal.) for changing an effective field of view (FOV) in order not to sacrifice high accuracy. The simulation compares the root mean square (RMS) of the foveation noise between uniform and non-uniform resolutions, when a resolution level and a FOV level are changed, respectively. Experimental results show that the proposed method is consistent with the wide-angle space-variant image by the fovea sensor, i.e., it does not sacrifice high accuracy in the central FOV.
UR - http://www.scopus.com/inward/record.url?scp=70349274454&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=70349274454&partnerID=8YFLogxK
U2 - 10.1109/ROBIO.2009.4913201
DO - 10.1109/ROBIO.2009.4913201
M3 - Conference contribution
AN - SCOPUS:70349274454
SN - 9781424426799
T3 - 2008 IEEE International Conference on Robotics and Biomimetics, ROBIO 2008
SP - 1377
EP - 1382
BT - 2008 IEEE International Conference on Robotics and Biomimetics, ROBIO 2008
PB - IEEE Computer Society
T2 - 2008 IEEE International Conference on Robotics and Biomimetics, ROBIO 2008
Y2 - 21 February 2009 through 26 February 2009
ER -