Sung Bum Pan

Unified systolic arrays for computation of the DCT/DST/DHT

We propose unified systolic arrays for computation of the one-dimensional (1-D) and two-dimensional (2-D) discrete cosine transform/discrete sine transform/discrete Hartley transform (DCT/DST/DHT). By decomposing the transforms into even- and odd-numbered frequency samples, the proposed architecture computes the 1-D DCT/DST/DHT. Compared to the conventional methods, the proposed systolic arrays exhibit advantages in terms of the number of PEs and latency. We generalize the proposed structure for computation of the 2-D DCT/DST/DHT. The unified systolic arrays can be employed for computation of the inverse DCT/DST/DHT (IDCT/IDST/IDHT).

Eye and face detection using SVM

In a real-time automatic face recognition system, accurate face detection plays a very important part in recognition performance. We use color, edge, and binary information to detect eye pair candidate regions from an input image, then extract the face candidate region with the detected eye pair. We verify both the eye pair candidate region and the face candidate region using support vector machines (SVM). It is possible to perform fast and reliable face detection because we can protect false detection through this verification process. Experimental results demonstrate that the proposed approach shows an excellent face detection performance of over 99.2%.

VLSI architectures for block matching algorithms using systolic arrays

We investigate hardware implementation of block matching algorithms (BMAs) for motion estimation of moving sequences. Using systolic arrays, we propose VLSI architectures for the two-stage BMA and full search (FS) BMA. The two-stage BMA using integral projections reduces greatly the computational complexity with its performance comparable to that of the FS BMA. The proposed hardware architectures for the two-stage BMA and FS BMA are faster than the conventional hardware architectures with lower hardware complexity. Also, the proposed architecture of the first stage of the two-stage BMA is modeled in VHDL and simulated. Simulation results show the functional validity of the proposed architecture.

An ultra-low memory fingerprint matching algorithm and its implementation on a 32-bit smart card

We propose an ultra-low memory fingerprint matching algorithm and implement it on a 32-bit smart card. We first evaluated both the number of instructions executed and memory requirement of each step of a typical fingerprint matching algorithm. Then, we developed a memory-efficient algorithm for the most memory-consuming step (alignment) by doing more computations under the restriction of the real-time requirement. Our experimental results show that the proposed algorithm can reduce the required memory space by a factor of 62 and can be executed in real-time on a 32-bit smart card.

Efficient Image Retrieval Using Adaptive Segmentation of HSV Color Space

This paper presents an efficient image retrieval system using adaptive segmentation of hue, saturation and value (HSV) color space. We classify the image into n number of areas based on different selected ranges of hue and value, then each area is partitioned into m number of segments based on the number of pixels it contains, the area which has more pixels will be partitioned into more segments and the area which has less pixels will be partitioned into less number of segments. This is used as a feature vector. Retrieval system outputs the image with a high matching factor. A small demonstration system has been tested and shows superior performance compared with the simple color based retrieval systems.

Hiding biometric data for secure transmission

In this paper, we describe biometric watermarking techniques for secure user verification on the remote, multimodal biometric system employing both fingerprint and face information, and compare their effects on verification accuracy quantitatively. To hide biometric data with watermarking techniques, we first consider possible two scenarios. In the Scenario 1, we use a fingerprint image as a cover work and hide facial features into it. On the contrary, we hide fingerprint features into a facial image in the Scenario 2. Based on the experimental results, we confirm that the Scenario 2 is superior to the Scenario 1 in terms of the verification accuracy of the watermarked image.

A Survey and Proposed Framework on the Soft Biometrics Technique for Human Identification in Intelligent Video Surveillance System

Biometrics verification can be efficiently used for intrusion detection and intruder identification in video surveillance systems. Biometrics techniques can be largely divided into traditional and the so-called soft biometrics. Whereas traditional biometrics deals with physical characteristics such as face features, eye iris, and fingerprints, soft biometrics is concerned with such information as gender, national origin, and height. Traditional biometrics is versatile and highly accurate. But it is very difficult to get traditional biometric data from a distance and without personal cooperation. Soft biometrics, although featuring less accuracy, can be used much more freely though. Recently, many researchers have been made on human identification using soft biometrics data collected from a distance. In this paper, we use both traditional and soft biometrics for human identification and propose a framework for solving such problems as lighting, occlusion, and shadowing.

Local and global feature extraction for face recognition

This paper proposes a new feature extraction method for face recognition. The proposed method is based on Local Feature Analysis (LFA). LFA is known as a local method for face recognition since it constructs kernels which detect local structures of a face. It, however, addresses only image representation and has a problem for recognition. In the paper, we point out the problem of LFA and propose a new feature extraction method by modifying LFA. Our method consists of three steps. After extracting local structures using LFA, we construct a subset of kernels, which is efficient for recognition. Then we combine the local structures to represent them in a more compact form. This results in new bases which have compromised aspects between kernels of LFA and eigenfaces for face images. Through face recognition experiments, we verify the efficiency of our method.

A Study on the Korean Banknote Recognition Using RGB and UV Information

Since the utilization of finance self-service is getting increment, bank and financial institutions has provided various services using automatic banking systems. For better efficiency of utilization of automatic banking system, banknote recognition, performing banknote classification and counterfeit detection, is getting more important. This paper used color and UV information of bankonte for banknote recognition. We have improved the accuracy of banknote classification by classify the candidate of the kind of banknote and then applying size information of the banknote. Counterfeit detection is performed to comparing UV information of reference and input image after banknote classification. Our experimental results show that the performance of banknote classification and counterfeit detection are 99.1% and 98.3%.

Access control system with high level security using fingerprints

Biometric based applications guarantee for resolving numerous security hazards. As a method of preserving of privacy and the security of sensitive information, biometrics has been studied and used for the past few decades. Fingerprint is one of the most widely used biometrics. A number of fingerprint verification approaches have been proposed until now. However, fingerprint images acquired using current fingerprint input devices that have small field of view are from just very limited areas of whole fingertips. Therefore, essential information required to distinguish fingerprints could be missed, or extracted falsely. The limited and somewhat distorted information are detected from them, which might reduce the accuracy of fingerprint verification systems. In the systems that verify the identity of two fingerprints using fingerprint features, it is critical to extract the correct feature information. In order to deal with these problems, compensation of imperfect information can be performed using multiple impressions of enrollees fingerprints. In this paper, additional three fingerprint images are used in enrollment phase of fingerprint verification system. Our experiments using FVC 2002 databases show that the enrollment using multiple impressions improves the performance of the whole fingerprint verification system.