• The Journal of the Korea Contents Association
• /
• v.16 no.10
• /
• pp.514-521
• /
• 2016

The High Efficiency Video Coding (HEVC) standard provides superior coding efficiency by utilizing highly flexible block structure and more diverse coding modes. However, rate-distortion optimization (RDO) process for the decision of optimal block size and prediction mode requires excessive computational complexity. To alleviate the computation load, this paper proposes a new moment-based fast CU size decision algorithm for intra coding in HEVC. In the proposed method, moment values are computed in each CU block to estimate the texture complexity of the block from which the decision on an additional CU splitting procedure is performed. Unlike conventional methods which are mostly variance-based approaches, the proposed method incorporates the third-order moments of the CU block in the design of the fast CU size decision algorithm, which enables an elaborate classification of CU types and thus improves the RD-performance of the fast algorithm. Experimental results show that the proposed method saves 32% encoding time with 1.1% increase of BD-rate compared to HM-10.0, and 4.2% decrease of BD-rate compared to the conventional variance-based fast algorithm.

• Journal of Korea Multimedia Society
• /
• v.3 no.1
• /
• pp.23-33
• /
• 2000

In this paper, we describe an algorithm which can extract human faces with natural pose from complex backgrounds. This method basically adopts the concept that facial region has the nearly same gray level for all pixels within appropriately scaled blocks. Based on the idea, we develop a hierarchial process that first, a block image data with pyramid structure of input image is generated, and some candidate regions for facial regions in the block image are Quickly determined, then finally the detailed facial features; organs are decided. To find the features easily, we introduce a local gray level transform which emphasizes dark and small regions, and estimate the geometrical triangle constraints among the facial features. The merit of our method is that we can be freed from the parameter assignment problem since the algorithm utilize a simple brightness computation, consequently robust systems not being depended on specific parameter values can be easily constructed.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.17 no.4
• /
• pp.71-76
• /
• 2017

Visual perception is part of a field of study that examines the analysis and recognition of objects. There have been numerous studies on this topic, including research on the physiological and cognitive aspects of the visual system. Visual perception theory was established through such research efforts and the potential for its application in each field is being investigated. Mobile displays are a representative example of media in the new age. Therefore, this study is based on the understanding that research on visual perception must stress the importance of useful visual cognition. This study used displays to conduct legibility tests based on illumination intensity and contrast ratio. Two conditions relevant to legibility were tested 1) illumination intensity environment and 2) two luminance conditions proposed as measures to improve readability in existing mobile displays. The results of this study were analysed to determine the degree of legibility based on illumination intensity and contrast ratio, and measures for improving legibility were proposed.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.36 no.8C
• /
• pp.505-515
• /
• 2011

This paper describes a normalization method of Haar-like features used for object detection. Previous method which performs variance normalization on Haar-like features requires a lot of calculations, since it uses an additional integral image for calculating the standard deviation of intensities of pixels in a candidate window and increases possibility of false detection in the area where variance of brightness is small. The proposed normalization method can be performed much faster than the previous method by not using additional integral image and classifiers which are trained with the proposed normalization method show robust performance in various lighting conditions. Experimental result shows that the object detector which uses the proposed method is 26% faster than the one which uses the previous method. Detection rate is also improved by 5% without increasing false alarm rate and 45% for the samples whose brightness varies significantly.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.40 no.6
• /
• pp.149-156
• /
• 2003

This paper proposes a face recognition method based on modified Otsu binarization, Hu moment and linear discriminant analysis (LDA). Proposed method is robust to brightness, contrast, scale, rotation, and translation changes. Modified Otsu binarization can make binary images that have the invariant characteristic in brightness and contrast changes. From edge and multi-level binary images obtained by the threshold method, we compute the 17 dimensional Hu moment and then extract feature vector using LDA algorithm. Especially, our face recognition system is robust to scale, rotation, and translation changes because of using Hu moment. Experimental results showed that our method had almost a superior performance compared with the conventional well-known principal component analysis (PCA) and the method combined PCA and LDA in the perspective of brightness, contrast, scale, rotation, and translation changes with Olivetti Research Laboratory (ORL) database and the AR database.

• Korean Journal of Cognitive Science
• /
• v.18 no.3
• /
• pp.223-244
• /
• 2007

Two well known mis-localization effects, flash-lag effect (FLE) and the Pulfrich effect, have similar phenomena and theoretical explanations. In order to directly compare two phenomena, thereby examining the possibility that two phenomena have a common mechanism, the magnitudes of two effects were measured under the same experimental settings and stimuli. The Pulfrich depth was measured from each of four different brightness ratios of two moving stimuli, each of which was projected to each eye. The magnitude of FLE was measured from each of five different brightness levels, which were the same levels used in the Pulfrich experiment. The Pulfrich depth was increased with the increase of brightness ratio, whereas similar pattern was not found in the magnitude of FLE. Furthermore, actual Pulfrich depths were greatly different from those predicted from the difference of two FLEs. These results suggest that two phenomena may not have a common mechanism.

• Journal of the Korea Society of Computer and Information
• /
• v.10 no.1 s.33
• /
• pp.67-72
• /
• 2005

Most of image binarization algorithms analyzes the intensity distribution using the histogram for the determination of threshold value. When the intensity difference between the foreground object and the background is great, the histogram shows the tendency to be bimodal and the selection of the histogram valley as the threshold value shows the good result. On the other side. when the intensity difference is not great and the histogram doesn't show the bimodal property, the histogram analysis doesn't support the selection of the proper threshold value. This Paper Proposed the novel binarization method that applies the fuzzy membership function to each color value on the RGB color model and, by using the operation results, separates the features having the great readability from the background. The proposed method prevents the loss of information incurred by the gray scale conversion by using the RGB color model and extracts effectively the readable features by using the fuzzy inference Compared with the traditional binarization methods, the proposed method is able to remove the majority of noise areas and show the improved results on the image of transport containers , etc.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.18 no.4
• /
• pp.490-498
• /
• 1993

A variety of methods for measuring the velocity from an image sequence use the relationship between the spatial and temporal gradients of image brightness function. In most situations, an additional constraint is required because the velocity is not determined uniquely by a above relationship. Horn and Schunch proposed a constraint that the velocity field should vary smoothly over the image. This requirement, however, forces the velocity field to vary smoothly even across motion boundaries. To complement this probe, Nagel introduced and 'oriented smoothness' constraint which restricts variations of velocity field only in directions with small or no variation of image brightness function. On the other hand, Paquin and Dubois proposed a different type of constraint that the velocity is constant in a small area of image. But, this constraint also creates difficulties at motion boundaries which large variations in velocity field often occur. We propose the method to overcome these difficulties by utilizing the information of discontinuities in image brightness function, and present the experimental results.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.14 no.3
• /
• pp.310-315
• /
• 2004

Image taken in the night can be low-contrast images because of poor environment and image transmission. We propose an algorithm that improves the acquired low-contrast image. MPEG-2 separates chrominance and illuminance, and compresses respectively because human vision is more sensitive to luminance. We extracted illumination and used K-means algorithm to find a proper crossover point automatically. We used K-means algorithm in the viewpoint that the problem of crossover point selection can be considered as the two-category classification problem. We divided an image into two subimages using the crossover point, and applied the histogram equalization method respectively. We used the index of fuzziness to evaluate the degree of improvement. We compare the results of the proposed method with those of other methods.

• Journal of Broadcast Engineering
• /
• v.24 no.6
• /
• pp.1113-1121
• /
• 2019

In this paper, we propose the face anti-spoofing method based on combination of luminance and chrominance with convolutional neural networks. The proposed method extracts luminance and chrominance features independently from live and fake faces by using stacked convolutional neural networks and auxiliary networks. Unlike previous methods, an attention module has been adopted to adaptively combine extracted features instead of simply concatenating them. In addition, we propose a new loss function, called the contrast loss, to learn the classifier more efficiently. Specifically, the contrast loss improves the discriminative power of the features by maximizing the distance of the inter-class features while minimizing that of the intra-class features. Experimental results demonstrate that our method achieves the significant improvement for face anti-spoofing compared to existing methods.