• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제12권7호
• /
• pp.3172-3193
• /
• 2018

Recently, several studies have shown that linear representation based approaches are very effective and efficient for image classification. One of these linear-representation-based approaches is the Collaborative representation (CR) method. The existing algorithms based on CR have two major problems that degrade their classification performance. First problem arises due to the limited number of available training samples. The large variations, caused by illumintion and expression changes, among query and training samples leads to poor classification performance. Second problem occurs when an image is partially noised (contiguous occlusion), as some part of the given image become corrupt the classification performance also degrades. We aim to extend the collaborative representation framework under limited training samples face recognition problem. Our proposed solution will generate virtual samples and intra-class variations from training data to model the variations effectively between query and training samples. For robust classification, the image patches have been utilized to compute representation to address partial occlusion as it leads to more accurate classification results. The proposed method computes representation based on local regions in the images as opposed to CR, which computes representation based on global solution involving entire images. Furthermore, the proposed solution also integrates the locality structure into CR, using Euclidian distance between the query and training samples. Intuitively, if the query sample can be represented by selecting its nearest neighbours, lie on a same linear subspace then the resulting representation will be more discriminate and accurately classify the query sample. Hence our proposed framework model the limited sample face recognition problem into sufficient training samples problem using virtual samples and intra-class variations, generated from training samples that will result in improved classification accuracy as evident from experimental results. Moreover, it compute representation based on local image patches for robust classification and is expected to greatly increase the classification performance for face recognition task.

• 대한전자공학회논문지SP
• /
• 제45권4호
• /
• pp.1-9
• /
• 2008

본 논문은 DC 영상을 이용하여 기하학적 변환에 강인하며 고속처리가 가능한 워터마킹 알고리즘을 제안한다. 제안하는 방법은 JPEG 및 MPEG 영상압축 기술에 널리 사용되는 $8{\times}8$ 블록 DCT를 기반으로 하며, DC 성분만으로 구성된 DC 영상에 DWT 수행 후 웨이블릿 계수에 워터마크를 삽입한다. DC 성분을 이용한 워터마크 삽입으로 야기될 수 있는 화질 열화를 극복하기 위하여 DWT의 서브밴드 마다 워터마크 삽입 강도와 삽입량을 차별화한다. 제안하는 알고리즘은 $8{\times}8$ 블록 DCT 영역의 부분적인 복호화(즉, DC 영상)를 수행함으로써 실시간 워터마크 삽입 및 추출을 위한 고속처리가 가능하다. 실험결과로부터 제안한 워터마킹 방법이 기존의 DC 성분을 이용한 알고리즘과 비교하여 82%의 계산시간을 단축하였으며 다양한 기하학적 공격과 JPEG 압축 등의 기타 공격에 강인한 것을 확인하였다.

• 한국통신학회논문지
• /
• 제33권11C호
• /
• pp.925-931
• /
• 2008

동영상 표준안 H.264/AVC에서 가변 블록 모드 알고리즘은 압축 성능을 향상시키나 다양한 블록 모드와 모드 결정을 위해 방대한 계산을 필요로 한다. 한편, 결정된 인터 블록 모드들은 블록 영상의 복잡도에 의존하여 복잡한 매크로블록일수록 블록 크기는 더 작아지고 있다. 본 논문은 고속 인터 블록 모드 결정 알고리즘을 제안한다. 제안된 알고리즘은 영상 복잡도를 이용해 유효 인터 블록 모드를 결정 가능성이 큰 블록 모드로 제한하고, 유효블록 모드들에서만 움직임 추정 및 율-왜곡 최적화를 수행한다. 거기에 $16{\times}16$ 블록 모드만을 갖는 유효 블록 모드들에는 고속 움직임 추정 알고리즘인 PDE (partial difference elimination)를 적용시키고 있다. 제안된 알고리즘의 성능 평가를 위해 참조 소프트웨어 JM 9.5에서 화질, 비트율, 부호화 시간이 기존 알고리즘과 비교되었다. 실험 결과는 제안된 알고리즘이 평균 움직임 추정 시간의 약 24.12%를 감소시키며 화질과 비트량을 각각 -0.02dB와 -0.12% 정도로 유지시킬 수 있는 것을 보여주었다.

• 마케팅과학연구
• /
• 제18권4호
• /
• pp.307-331
• /
• 2008

The CEO and the Corporate Image is considered very important in the aspect of marketing. The fact that CEO image itself influences the company or value of the product directly and indirectly has been verified through many cases. Recently, the differentiation of products and services between companies became difficult because the disparity in technique between companies retrenched. As a result, the rate of people who decide to purchase or invest their money based on the corporate image or reputation has been increased. Also in the knowledge society like today, the talented employees are the company's customer and the company's necessity for managing those brains of marketing perspective on how to satisfy and attract the customers is being embossed. The Fashion industry is one of the most value-added industry and in those value-added businesses, the most important factor is the human resources' knowledge power. However the study of the relationships among the CEO image, the corporate image and employment brand value in fashion industry has not been carried out yet. This research considers that dynamic relationship exists among the CEO image, corporate image and employment brand value that affects a company's main goal of pursuing benefits and intends to investigate the relationships of the three concepts. The specific purposes of this study were, 1) to analyze the impact of CEO image on a corporate image, 2) to analyze the impact of corporate image on employment brand value, 3) to analyze the impact of CEO image on employment brand value, 4) to analyze whether corporate image plays a mediating role in the relationship between CEO image and employment brand value or not. A survey design with a structured questionnaire was employed for this research. A convenience sample of 398 subjects was selected from two groups, which are university students majoring in fashion and practitioners working in fashion industry. For the data analysis, descriptive statistic (i.e., frequency, percentage), factor analysis, and multiple regression analysis were used by utilizing SPSS 12.0 for Windows program. The results for this research are as follows, first, the study of the impact of CEO image (i.e., Managerial Competence, Reliability/Leadership, Personal Attractiveness) on corporate image (i.e., Product Image, Corporate Social Responsibility Image, Corporate Cultural Image) brought conclusion that the CEO image generally affected the corporate image in fashion industry. Managerial Competence and Reliability/Leadership affected Product Image, Corporate Social Responsibility Image and Corporate Cultural Image. However, while CEO's Personal Attractiveness affected Product Image and Corporate Social Responsibility Image, it did not affect Corporate Cultural Image. Second, the study of the impact of corporate image on employment brand value brought conclusion that corporate image (i.e., Product Image, Corporate Social Responsibility Image, Corporate Cultural Image) affected employment brand value. Corporate Cultural Image affected employment brand value the most and then the Corporate Social Responsibility Image and Product Image. Third, the study of the impact of CEO image on employment brand value brought conclusion that CEO image (i.e., Managerial Competence, Reliability/Leadership, Personal Attractiveness) affected the employment brand value. CEO's Reliability/Leadership affected the employment brand value the most and then CEO's Personal Attractiveness and CEO's Managerial Competence. Forth, the study examined whether corporate image plays a mediating role in relationship of CEO image and employment brand value and concluded that it does. Corporate image played a full mediating role between CEO's Managerial Competence and employment brand value while it played a partial mediating role between CEO's Reliability/Leadership and CEO's Personal Attractiveness. This study is meaningful in a sense that it examines the relationship among the CEO image, corporate image and employment brand value which has not been carried out yet in fashion industry. It will ultimately contribute to the success of a fashion company by providing useful information of establishing strategies for managing proper the CEO and the corporate image to the fashion company and operating the talented employees.

• Progress in Superconductivity
• /
• 제11권1호
• /
• pp.72-77
• /
• 2009

Low temperature scanning laser microscopy (LTSLM) can be used for a two-dimensional display of bolometric response arising from the localized excitation of a sample by the focused laser beam. In this study, the distribution of critical temperature ($T_c$) and critical current density ($J_c$) in YBCO coated conductor were analyzed using LTSLM. For improving the temperature stability, we have modified the system into a double-shielding type. Through the modification, the temperature stability was successfully improved from ${\pm}10mK\;to\;{\pm}2mK$. The superconducting properties of YBCO coated conductors were measured for the sample of a narrow bridge type using wet etching process. The spatial non-uniformity of the ac voltage response, ${\delta}V(x)$, which is proportional to ${\partial}\rho(x,J_B)/{\partial}T$ in the transition temperature region could be observed and displayed in a two-dimensional image.

• Journal of Electrical Engineering and information Science
• /
• 제3권2호
• /
• pp.211-221
• /
• 1998

It is becoming apparent that stereo matching algorithms need much information from high level cognitive processes. Otherwise, conventional algorithms based on bottom-up control alone are susceptible to local minima. We introduce a system that consists of two levels. A lower level, using a usual matching method, is based upon the local neighborhood and a second level, that can integrate the partial information, is aimed at contextual matching. Conceptually, the introduction of bottom-up and top-down feedback loop to the usual matching algorithm improves the overall performance. For this purpose, we model the image attributes using a Markov random field (MRF) and thereupon derive a maximum a posteriori (MAP) estimate. The energy equation, corresponding to the estimate, efficiently represents the natural constraints such as occlusion and the partial informations from the other levels. In addition to recognition, we derive a training method that can determine the system informations from the other levels. In addition to recognition, we derive a training method that can determine the system parameters automatically. As an experiment, we test the algorithms using random dot stereograms (RDS) as well as natural scenes. It is proven that the overall recognition error is drastically reduced by the introduction of contextual matching.

• 한국광학회지
• /
• 제19권2호
• /
• pp.103-107
• /
• 2008

기존 OCT 시스템의 Michelson 간섭계와 다르게 간섭신호가 프로브 끝단에서 유도되어 기준신호와 샘플신호가 동일한 경로를 통해 진행하는 공통경로 OCT 시스템을 제작하였다. 최적의 기준신호가 생성되도록 하기 위해 부분반사 프로브(Probe)의 굴절률을 최적화하였고 생체 샘플의 이미지 획득이 가능하도록 프로브의 끝단에서 초점이 맺히도록 볼록 렌즈형 프로브를 제작하였다. 이를 통해서 실제 샘플의 2차원 이미지의 획득이 가능하였다.

• 전자공학회논문지S
• /
• 제35S권1호
• /
• pp.130-140
• /
• 1998

This paper presents a method to estimate subpixel accuracy motion vectors using a mathermatical model withoug interpolation. the proposed method decides the coefficients of mathematical model, which represents the motion vector which is achieved by full search. And then the proposed method estimates subpixel accuracy motion vector from achieved mathematical model. Step by step mathematical models such as type 1, type 2, type 3, modified bype 2, modified type 3, and Partial Interpolation type 3 are presented. In type 1, quadratic polynomial, which has 9 unknown coefficients and models the 3by 3 pixel plane, is used to get the subpixel accuracy motion vectors by inverse matrix solution. In type 2 and 3, each quadratic polynomial which is simplified from type 1 has 5 and 6 unknown coefficients and is used by least square solution. Modified type 2 and modified type 3 are enhanced models by weighting only 5 pixels out of 9. P.I. type 3 is more accurate method by partial interpolation around subpixel which isachieved by type 3. LThese simulation results show that the more delicate model has the better performance and modified models which are simplified have excellent performance with reduced computational complexity.

• Clinics in Shoulder and Elbow
• /
• 제9권1호
• /
• pp.68-72
• /
• 2006

Purpose: To demonstrate arthroscopically the presence of bilateral shoulder "mirror lesions" due to overhead work. Materials and Methods: A cross-sectional study was performed in a historical cohort of 10 male tire plant workers who underwent MRI studies of both shoulders. Bilateral intraarticular shoulder pathology was diagnosed by magnetic resonance imaging and confirmed by arthroscopy. "Mirror lesions" were defined as similar pathology between an individual patient's two shoulders. Results: We report on ten patients who had bilateral "mirror lesions." The most common mirror lesions were tears of the rotator cuff. Most patients (7 of 10) had more than one mirror lesion in their shoulders. There were two bilateral full thickness tears involving the supraspinatus and infraspinatus, four bilateral partial thickness supraspinatus tears, six bilateral partial or upper corner lesion complete subscapularis tears, and one bilateral complete supraspinatus tear. Conclusion: Occupational overhead work with simultaneous use of both arms is associated with bilateral shoulder lesions verified by arthroscopy. By demonstrating the work-related bilateral shoulder involvement, this study suggests an etiopathogenesis for these lesions and provides rational for developing worksite prevention strategies.

• Journal of the Optical Society of Korea
• /
• 제10권1호
• /
• pp.16-22
• /
• 2006

Wavefront sensing using a Shack-Hartmann sensor has been widely used for estimating wavefront errors or distortions. The sensor combines the local slopes, which are estimated from the centroids of each lenslet image, to give the overall wavefront reconstruction. It was previously shown that the pupil-plane irradiance profile effects the centroid estimation. Furthermore, a previous study reported that the reconstructed wavefront from a planar wavefront with a Gaussian pupil irradiance profile contains large focus and spherical aberration terms when there is a focus error. However, it has not been reported yet how seriously the pupil irradiance profiles, which can occur in practical applications, effect the sensing errors. This paper considered two cases when the irradiance profiles are not uniform: 1) when the light source is Gaussian and 2) when there is a partial interference due to a double reflection by a beam splitting element. The images formed by a Shack-Hartmann sensor were simulated through fast Fourier transform and were then supposed to be detected by a noiseless CCD camera. The simulations found that sensing errors, due to the Gaussian irradiance profile and the partial interference, were found to be smaller than RMS ${\lambda}/50$ when ${\lambda}$ is $0.6328\;{\mu}m$, which can be ignored in most practical cases where the reference and test beams have the same irradiance profiles.