An efficient scale and rotation invariant 2-D object recognition
method
H. Lee, H. Kwon, and H. Hwang. Speech, Image Processing and Neural Networks, 1994. Proceedings,
ISSIPNN '94., 1994 International Symposium on, 2, page 405--408. IEEE Computer Society, (1994)
DOI: 10.1109/SIPNN.1994.344882
Abstract
This paper proposes an efficient scale and rotation invariant 2-D
object recognition method using Complex-Log Mapping (CLM) and Translation
Invariant Neural Network (TINN). CLM is known as very useful transform
for extracting scale and rotation invariant features. However, the
results are given in a wrap-around translated form, which requires
subsequent wrap-translation invariant recognition steps. Recently,
a new method using an augmented second order neural network (SONN)
was introduced as a solution. It requires, however, a connection
complexity O(n2) for input feature extraction which is too high to
be implemented. In this paper, we propose a method reducing the connection
complexity to O(n*log(n)) by using TINN. Experimental results show
that the recognition performance of the proposed method is almost
the same as that of SONN while its network size is significantly
reduced
%0 Conference Paper
%1 Lee1994
%A Lee, Heung-Ho
%A Kwon, Hee-Yong
%A Hwang, Hee-Yeung
%B Speech, Image Processing and Neural Networks, 1994. Proceedings,
ISSIPNN '94., 1994 International Symposium on
%D 1994
%I IEEE Computer Society
%K 2D complex-log complexity, computational connection extraction, feature image invariant mapping, nets network neural object recognition, rotation scale sequences, translation
%P 405--408
%R 10.1109/SIPNN.1994.344882
%T An efficient scale and rotation invariant 2-D object recognition
method
%V 2
%X This paper proposes an efficient scale and rotation invariant 2-D
object recognition method using Complex-Log Mapping (CLM) and Translation
Invariant Neural Network (TINN). CLM is known as very useful transform
for extracting scale and rotation invariant features. However, the
results are given in a wrap-around translated form, which requires
subsequent wrap-translation invariant recognition steps. Recently,
a new method using an augmented second order neural network (SONN)
was introduced as a solution. It requires, however, a connection
complexity O(n2) for input feature extraction which is too high to
be implemented. In this paper, we propose a method reducing the connection
complexity to O(n*log(n)) by using TINN. Experimental results show
that the recognition performance of the proposed method is almost
the same as that of SONN while its network size is significantly
reduced
@inproceedings{Lee1994,
abstract = {This paper proposes an efficient scale and rotation invariant 2-D
object recognition method using Complex-Log Mapping (CLM) and Translation
Invariant Neural Network (TINN). CLM is known as very useful transform
for extracting scale and rotation invariant features. However, the
results are given in a wrap-around translated form, which requires
subsequent wrap-translation invariant recognition steps. Recently,
a new method using an augmented second order neural network (SONN)
was introduced as a solution. It requires, however, a connection
complexity O(n2) for input feature extraction which is too high to
be implemented. In this paper, we propose a method reducing the connection
complexity to O(n*log(n)) by using TINN. Experimental results show
that the recognition performance of the proposed method is almost
the same as that of SONN while its network size is significantly
reduced},
added-at = {2011-03-27T19:35:34.000+0200},
author = {Lee, Heung-Ho and Kwon, Hee-Yong and Hwang, Hee-Yeung},
biburl = {https://www.bibsonomy.org/bibtex/296278544f0f133af2aca2254d948a7a1/cocus},
booktitle = {Speech, Image Processing and Neural Networks, 1994. Proceedings,
ISSIPNN '94., 1994 International Symposium on},
booktitleaddon = {April 13--16, 1994},
doi = {10.1109/SIPNN.1994.344882},
file = {:./00344882.pdf:PDF},
interhash = {c85a82b75222428958b60be762e606ee},
intrahash = {96278544f0f133af2aca2254d948a7a1},
keywords = {2D complex-log complexity, computational connection extraction, feature image invariant mapping, nets network neural object recognition, rotation scale sequences, translation},
location = {#ieeeaddr#},
pages = {405--408},
publisher = {{IEEE} Computer Society},
timestamp = {2011-03-27T19:35:41.000+0200},
title = {An efficient scale and rotation invariant 2-D object recognition
method},
volume = 2,
year = 1994
}