An Ear Recognition Method Based on Rotation Invariant Transformed DCT

Abstract

Human recognition systems have gained great importance recently in a wide range of applications like access, control, criminal investigation and border security. Ear is an emerging biometric which has rich and stable structure and can potentially be implemented reliably and cost efficiently. Thus human ear recognition has been researched widely and made greatly progress. High recognition rates which are reported in most existing methods can be reached only under closely controlled conditions. Actually a slight amount of rotation and translation which is inescapable would be injurious for system performance. In this paper, a method that uses a transformed type of DCT is implemented to extract meaningful features from ear images. This algorithm is quite robust to ear rotation, translation and illumination. The proposed method is experimented on two popular databases, i.e. USTB II and IIT Delhi II, which achieves significant improvement in the performance in comparison to other methods with good efficiency based on LBP, DSIFT and Gabor. Also because of considering only important coefficients, this method is faster compared to other methods.

Enhancement Ear-based Biometric System Using a Modified AdaBoost Method

“…This section presents related work using ear recognition. Hourali, F., and Gharravi present a method that uses a transformed type of DCT to extract meaningful features from ear images 3 .…”

An Accurate Facial Component Detection Using Gabor Filter

“…Rubel Biswas et al [12] have proposed a system for detection and recognition of iris using Hough transformation for object detection and SVM classifiers are used for training. Fatemeh Hourali et al [13] have been discussed about ear recognition using Invariant transformed DCT.…”

A human vision based system for biometric images recognition

“…Based on the significant separation property of the LSM which assumes that different inputs to a pool of neurons; which represents the liquid or reservoir of the LSM, should cause different neuron responses; and similar inputs should produce same responses [24], the successful LSM results obtained dealing with analogue signals as inputs and finally the experimental results in biology and neurophysiology which proved that the visual system analyses inputs in several spatial resolution scales; which motivated the use of spatial frequency preprocessing of images [25]- [27] in computer vision and pattern recognition, we propose a universal human based biometric identification system using the LSM with its biologically inspired parameters and structure with a new image encoding method using, rather than pixels, an analogue signal obtained by mapping a frequency filtered image. To increase furthermore the efficiency of the proposed system, every input image is splitted in 16 fragments.…”