• Journal of Korea Multimedia Society
• /
• v.15 no.6
• /
• pp.761-769
• /
• 2012

In this paper, a setting method of initial cluster centers and color image segmentation using superpixels and Fuzzy C-means(FCM) algorithm is proposed. Generally, the FCM can be widely used to segment color images, and an element is assigned to any cluster with each membership values in the FCM. However the algorithm has a problem of local convergence by determining the initial cluster centers. So the selection of initial cluster centers is very important, we proposed an effective method to determine the initial cluster centers using superpixels. The superpixels can be obtained by grouping of some pixels having similar characteristics from original image, and it is projected $La^*b^*$ feature space to obtain the initial cluster centers. The proposed method can be speeded up because number of superpixels are extremely smaller than pixels of original image. To evaluate the proposed method, several color images are used for computer simulation, and we know that the proposed method is superior to the conventional algorithm by the experimental results.

• The Journal of the Acoustical Society of Korea
• /
• v.25 no.1
• /
• pp.14-20
• /
• 2006

In this paper. a novel technique to set up the initial weights in BSS of delayed mixtures is proposed. After analyzing Eigendecomposition for the correlation matrix of mixing data. the initial weights are set from the Eigenvectors ith delay information. The Proposed setting of initial weighting method for conventional FDICA technique improved the separation Performance. The computer simulation shows that the Proposed method achieves the improved SIR and faster convergence speed of learning curve.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.26 no.5B
• /
• pp.582-591
• /
• 2001

터보 부호는 반복적인 복호 알고리즘을 사용함으로써 가산성 백색 가우시안 잡음(AWGN) 채널 환경에서 Shannon 한계에 가까운 성능을 보이는 오류정정 방식으로 제안되었으나, 반복 연산량에 따른 복호 지연과 인터리버에 따른 지연에 의해 실시간 처리의 어려움이라는 문제점을 안고 있다. 본 논문에서는 터보 부호의 성능을 저하시키지 않는 범위에서 적절한 초기 임계값 설정에 따라 불필요한 반복 복호 횟수를 줄일 수 있는 터보 복호기 구조를 제안한다. 적절한 초기 임계값 설정은 LLR(Log-Likelihood Ratio)값의 평균값과 분산, 복호기의 출력에 대한 BER에 근거하여 여러 번의 모의 실험을 통해서 최적의 값으로 결정된다. 제안한 방식은 초기 임계값을 적절히 선택하면 손실이 없는 범위 내에서 반복횟수를 감소시킴으로써 기존의 정해진 반복횟수로 인한 큰 복호 지연을 미연에 방지하고, 이에 따른 계산량 감소는 저전력의 효과도 가져온다. 성능 평가를 위해 BER = $10^{-6}$이내이고, 전송속도가 32kbps 이상인 IMT2000의 고속 데이터 전송 환경에서 모의 실험을 하였다. 실험 결과로 기존의 정해진 반복횟수를 갖는 터보 복호기에 비해 SNR 변동(0~3dB)에서 평균적으로 55~90% 정도의 감소된 반복횟수를 검증하였다.

• Communications for Statistical Applications and Methods
• /
• v.4 no.1
• /
• pp.91-99
• /
• 1997

본 연구에서는 신경망이론을 이용하여 시계열자료를 분석할 때 문제가 되고 있는 초기 가중값을 선정하는 방법을 제시하고자 한다. 기존의 연구에서 학습을 위한 초기 가중값의 결정은 난수에 의존하고 있다. 본 연구에서는 신경망학습의 효율적인 초기값을 선택하기 위하여 제어상자를 이용한다. 그리고 학습과정에서 가중값의 변화를 추적하고 적절한 가중값의 범위를 탐색하면서 새로운 초기값을 제어상자를 통하여 실시간으로 재설정하는 방법을 제시한다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2001.04b
• /
• pp.496-498
• /
• 2001

Snake 모델(active contour model)은 초기값을 설정해주면 자동으로 임의의 물체의 윤곽을 찾아내는 알고리즘으로 영상에서 특정 영역을 분할하여 할 때 많이 이용되고 있다. 본 논문에서는 칼라 영상에서 얼굴과 얼굴의 특징점을 찾는 방법으로 이 알고리즘을 적용한다. 특히, 주어진 영상의 RGB 값을 정규화(normalization) 해주는 전처리 과정을 통해 얼굴의 특징점 후보 영역을 얻어내는 초기 값을 설정해주어야 하는 과정을 생략해주고 보다 정확한 값을 얻을 수 있도록 구현한다. RGB 값을 이용한 정규화 과정을 적용한 방법과 적용하지 않은 방법을 구현한 결과를 비교해줌으로써, 정규화 과정을 거친 방법의 성능이 더 우수함을 보여준다.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2004.05b
• /
• pp.599-602
• /
• 2004

In this paper, an automatic setting method of control parameters for robot scorer is proposed. First we acquisited some local image lesions including robots and ball patch, and sampled the regions to RCB value. Edge operator is applied to get magnitude of gradient at each pixel. Some effective patch regions can be detected by magnitude of gradient, and YUV value at each pixel in patch lesions can be obtained. We can determine control parameters of robot soccer using luminance component of YUV. The proposed method is applied to robot soccer image to detect initial patch value and get control parameters adaptively in light variation.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.13 no.10
• /
• pp.2071-2078
• /
• 2009

The first frame is encoded in intra mode which generates a larger number of bits. In addition, the first frame is used for the inter mode encoding of the following frames. Thus the intial QP (Quantization Parameter) for the first frame affects the first frame as well as the following frames. Traditionally, the initial QP is determined among four constant values only depending on the bpp. In the case of low bit rate video coding, the initial QP value is fixed to 35 regardless of the output bandwidth. Although this initialization scheme is simple, yet it is not accurate enough. An accurate intial QP prediction scheme should not only depends on bpp but also on the complexity of the video sequence and the output bandwidth. In the proposed scheme, we use a linear model because there is a linear inverse proportional relationship between the output bandwidth and the optimal intial QP. Model parameters of the model are determined depending on the spatial complexity of the first frame. It is shown by experimental results that the new algorithm can predict the optimal initial QP more accurately and generate the PSNR performance better than that of the existing JM algorithm.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.16 no.9
• /
• pp.1947-1954
• /
• 2012

The first frame is encoded in intra mode which generates a larger number of bits. In addition, the first frame is used for the inter mode encoding of the following frames. Thus the initial QP for the first frame affects the first frame as well as the following frames. Traditionally, the initial QP is determined among four constant values only depending on the bpp. In the case of low bit rate video coding, the initial QP value is fixed to 40 regardless of the output bandwidth. Although this initialization scheme is simple, yet it is not accurate enough. An accurate initial QP prediction scheme should not only depends on bpp but also on the complexity of the video sequence and the output bandwidth. In the proposed scheme, we determine the initial QP according to the ratio of the first frame to the total bits allocated to a GOP. To estimate the QP of the allocated bits, Rate-QP model is used. It is shown by experimental results that the new algorithm can predict the optimal initial QP more accurately and generate the PSNR performance better than that of the existing JVT algorithm.

• The Korean Journal of Applied Statistics
• /
• v.32 no.5
• /
• pp.693-702
• /
• 2019

Deep learning has gained popularity for the classification and prediction task. Neural network layers become deeper as more data becomes available. Saturation is the phenomenon that the gradient of an activation function gets closer to 0 and can happen when the value of weight is too big. Increased importance has been placed on the issue of saturation which limits the ability of weight to learn. To resolve this problem, Glorot and Bengio (Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, 249-256, 2010) claimed that efficient neural network training is possible when data flows variously between layers. They argued that variance over the output of each layer and variance over input of each layer are equal. They proposed a method of initialization that the variance of the output of each layer and the variance of the input should be the same. In this paper, we propose a new method of establishing initialization by adopting truncated normal distribution and truncated cauchy distribution. We decide where to truncate the distribution while adapting the initialization method by Glorot and Bengio (2010). Variances are made over output and input equal that are then accomplished by setting variances equal to the variance of truncated distribution. It manipulates the distribution so that the initial values of weights would not grow so large and with values that simultaneously get close to zero. To compare the performance of our proposed method with existing methods, we conducted experiments on MNIST and CIFAR-10 data using DNN and CNN. Our proposed method outperformed existing methods in terms of accuracy.

• Journal of the Korean Association of Geographic Information Studies
• /
• v.4 no.4
• /
• pp.1-7
• /
• 2001

Remotely sensed data is the most fundamental data in acquiring the GIS informations, and may be analyzed to extract useful thematic information. Multi-spectral classification is one of the most often used methods of information extraction. The actual multi-spectral classification may be performed using either supervised or unsupervised approaches. This paper analyze the effect of assigning clever initial values to image classes on the performance of parallelepiped classification algorithm, which is one of the supervised classification algorithms. First, we investigate the effect on serial computing model, then expand it on MIMD(Multiple Instruction Multiple Data) parallel computing model. On serial computing model, the performance of the parallel pipe algorithm improved 2.4 times at most and, on MIMD parallel computing model the performance improved about 2.5 times as clever initial values are assigned to image class. Through computer simulation we find that initial values of image class greatly affect the performance of parallelepiped classification algorithms, and it can be improved greatly when classes on both serial computing model and MIMD parallel computation model.