• Proceedings of the Korea Information Processing Society Conference
• /
• 2000.10b
• /
• pp.887-890
• /
• 2000

본 연구에서는 내용기반 영상 데이터 검색을 위하여 변환 영역에서 위치 정보와 주파수 정보를 가지는 웨이블릿 성질을 이용하여 영상을 압축한 후에 저주파 성분에 의한 객체들의 특징을 추출하는 방안으로 Vector Quantization 을 이용한 class 별 영상 검색을 제시한다 내용기반 영상 검색의 주요특징들은 색상, 질감, 그리고 영상의 공간적인 특징을 고려한 특징 값 둥이 사용된다. 먼저 검색의 효율성을 높이기 위해 영상을 구성하는 특징 치 중에서 가장 빈도가 많은 class 부터 영상의 유사도를 검색한 후에 다음으로 영상을 구성하는 빈도가 큰 순서대로 DB 내에 저장되어 있는 영상과 비교를 하게 된다. DB내 영상 검색은 빈도수가 우선인 5개의 class를 기준으로 유사도를 측정해서 검색을 이룬다. 이러한 영상의 특징들을 어떻게 결합하고 특징 추출을 하느냐에 따라 검색의 효율성에 영향을 준다. 따라서 본 연구에서는 영상의 위치 정보와 주파수 정보를 가지는 웨이블릿 변환 후 얻어지는 저대역 부밴드에서의 공간적인 특성을 고려한 특징 값을 이용하여 Vector Quantization 알고리즘에 의해 정지영상의 객체 대표 특징들을 마르게 검색하고자 한다. 본 연구에서는 Haar Wavelet과 Vector Quantization 에서 색상과 질감의 가중치를 적용한 후 DB 에 저장된 영상과 유사도를 검색하는 방법을 취하고자 한다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 1997.11a
• /
• pp.55-62
• /
• 1997

본 논문에서는 윤곽선의 다각근사화시 발생하는 오차신호를 표본화하고 표본화된 신호를 부호화하기 위한 새로운 변환을 제안한다. 제안하는 방법은 다각근사화시 발생하는 윤곽선의 단편들이 완만한 곡선의 형태를 갖고 양 끝점이 0이라는 특징을 고려해서 르장드르 다항식에 기반한 새로운 변환을 제안한다. 제안한 방법은, 기존의 방법인 이산여현변환, 이산정현변환 등과 변환부호화 이득면에서 비교해 볼 때 우수한 결과를 얻었다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.10 no.4
• /
• pp.758-763
• /
• 2006

In various fields of basic science and engineering, in order to present signals and systems exactly and acquire useful information from spatial and timely changes, many researches have been processed. In these methods, the Fourier transform which represents signal as the combination of the frequency component has been applied to the most fields. But as transform not to consider time information, the Fourier transform has its limitations of application. To overcome this problem, a variety of methods including the wavelet transform have been proposed. As transform to represent signal by using the changing window, according to scale parameter in time-scale domain, the wavelet transform is capable of multiresolution analysis and defines various functions according to the application environments. In this paper, to detect features of signal we analyzed wavelet the spectrum by using the basis function of the fourier transform.

• Journal of Digital Contents Society
• /
• v.11 no.1
• /
• pp.17-24
• /
• 2010

The purpose of this study is to design and implement a kernel for the support vector machine(SVM) to improve the performance of face recognition. Local feature analysis(LFA) has been well known for its good performance. SVM kernel plays a limited role of mapping low dimensional face features to high dimensional feature space but the proposed kernel using LFA is designed for face recognition purpose. Because of the novel method that local face information is extracted from training set and combined into the kernel, this method is expected to apply to various object recognition/detection tasks. The experimental results shows its improved performance.

• Proceedings of the Korea Society for Industrial Systems Conference
• /
• 2009.05a
• /
• pp.138-143
• /
• 2009

본 논문에서는 감성 공간에서 mean shift clustering과 user feedback을 이용하여 영상 검색 결과를 개선하기 위한 시스템을 제안한다. 제안된 시스템은 사용자 인터페이스, 감성 공간 변환, 검색결과 순위 재지정(re-ranking)으로 구성된다. 사용자 인터페이스는 텍스트 형태의 질의 입력과 감성 어휘 선택에 따른 user feedback에 의해 개선된 검색결과를 보인다. 사용된 감성 어휘는 고바야시가 정의한 romantic, natural, casual, elegant, chic, classic, dandy, modern 등의 8개 어휘를 사용한다. 감성 공간 변환 단계에서는 입력된 질의에 따라 웹 영상 검색 엔진(Yahoo)에 의해 검색된 결과 영상들에 대해 컬러와 패턴정보의 특징을 추출하고, 이를 입력으로 하는 8개의 각 감성별 분류기에 의해 각 영상은 8차원 감성 공간으로의 특징 벡터로 변환된다. 이때 감성 공간으로 변환된 특징 벡터들은 mean shift clustering을 통해 군집화 되고, 그 결과로써 대표 클러스터를 찾게 된다. 검색결과 순위 재지정 단계에서는 user feedback 유무에 따라 대표 클러스터의 평균 벡터와 user feedback에 의해 생성된 사용자 감성 벡터에 의해 검색 결과를 개선할 수 있다. 이때 각 기준에 따라 유사도가 결정되고 검색결과 순위가 재지정 된다 제안된 시스템의 성능을 검증하기 위해 7개의 질의의 각 400장, 총 2,800장에 대한 Yahoo 검색 결과와 제안된 시스템을 개선된 검색 결과를 비교하였다.

• The KIPS Transactions:PartB
• /
• v.10B no.4
• /
• pp.459-464
• /
• 2003

This paper proposes the use of a wavelet based image transform algorithm in human iris recognition method and the effectiveness of this technique will be determined in preprocessing of extracting Iris image from the user´s eye obtained by imaging device such as CCD Camera or due to torsional rotation of the eye, and it also resolves the problem caused by invariant under translations and dilations due to tilt of the head. This technique values through the proposed translation-invariant wavelet transform algorithm rather than the conventional wavelet transform method. Therefore we extracted the best-matching iris feature values and compared the stored feature codes with the incoming data to identify the user. As result of our experimentation, this technique demonstrate the significant advantage over verification when it compares with other general types of wavelet algorithm in the measure of FAR & FRR.

• KIPS Transactions on Software and Data Engineering
• /
• v.4 no.4
• /
• pp.201-208
• /
• 2015

This paper proposes a method for detecting the direction indicator shown in the road surface efficiently from the black box system installed on the vehicle. In the proposed method, the direction indicators are detected by inverse perspective mapping(IPM) and bag of visual features(BOF)-based NN classifier. In order to apply the proposed method to real-time environments, the candidated regions of direction indicator in an image only performs IPM, and BOF-based NN is used for the classification of feature information from direction indicators. The results of applying the proposed method to the road surface direction indicators detection and recognition, the detection accuracy was presented at least about 89%, and the method presents a relatively high detection rate in the various road conditions. Thus it can be seen that the proposed method is applied to safe driving support systems available.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.2
• /
• pp.219-224
• /
• 2020

Binary classification has been broadly investigated in machine learning. In addition, binary classification can be easily extended to multi class problems. To successfully utilize machine learning methods for classification tasks, preprocessing and feature extraction steps are essential. These are important steps to improve their classification performances. In this paper, we propose a new learning method based on weighted least squares. In the weighted least squares, designing weights has a significant role. Due to this necessity, we also propose a new technique to obtain weights that can achieve feature transformation. Based on this weighting technique, we also propose a method to combine the learning and feature extraction processes together to perform both processes simultaneously in one step. The proposed method shows the promising performance on five UCI machine learning data sets.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.7 no.4
• /
• pp.598-603
• /
• 2003

This paper presents a sub-band feature extraction approach in which the feature extraction method in the individual frequency sub-bands is determined in terms of speech recognition accuracy. As in the multi-band paradigm, features are extracted independently in frequency sub-regions of the speech signal. Since the spectral shape is well structured in the low frequency region, the all pole model is effective for feature extraction. But, in the high frequency region, the nonparametric transform, discrete cosine transform is effective for the extraction of cepstrum. Using the sub-band specific feature extraction method, the linguistic information in the individual frequency sub-bands can be extracted effectively for automatic speech recognition. The validity of the proposed method is shown by comparing the results of speech recognition experiments for our method with those obtained using a full-band feature extraction method.

• Korean Journal of Remote Sensing
• /
• v.28 no.4
• /
• pp.357-367
• /
• 2012

This paper presents a two-stage methodology for anomaly detection from hyperspectral imagery that consists of transform-based feature extraction and selection, and computation of a local spatial auto-correlation statistic. First, principal component transform and 3D wavelet transform are applied to reduce redundant spectral information from hyperspectral imagery. Then feature selection based on global skewness and the portion of highly skewed sub-areas is followed to find optimal features for anomaly detection. Finally, a local indicator of spatial association (LISA) statistic is computed to account for both spectral and spatial information unlike traditional anomaly detection methodology based only on spectral information. An experiment using airborne CASI imagery is carried out to illustrate the applicability of the proposed anomaly detection methodology. From the experiments, anomaly detection based on the LISA statistic linked with the selection of optimal features outperformed both the traditional RX detector which uses only spectral information, and the case using major principal components with large eigen-values. The combination of low- and high-frequency components by 3D wavelet transform showed the best detection capability, compared with the case using optimal features selected from principal components.