• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.41 no.6
• /
• pp.61-71
• /
• 2004

Recently, various techniques for face detection are studied, but most of them still have problems on processing in real-time. Therefore, in this paper, we propose novel techniques for real-time detection of human faces in sequential images using motion and chroma information. First, background model is used to find a moving area. In this procmoving area. edure, intensity values for reference images are averaged, then skin-color are detected in We use HCr color-space model and adaptive threshold method for detection. Second, binary image labeling is applied to acquire candidate regions for faces. Candidates for mouth and eyes on a face are obtained using differences between green(G) and blue(B), intensity(I) and chroma-red(Cr) value. We also considered distances between eye points and mouth on a face. Experimental results show effectiveness of real-time detection for human faces in sequential images.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.22 no.4
• /
• pp.494-499
• /
• 2012

In this paper, a new method for detecting various objects that can be risks to safety navigation in sea environment is proposed. By analysing Infrared(IR) images obtained from various sea environments, we could find out that object regions include both horizontal and vertical direction edges while background regions of sea surface mainly include vertical direction edges. Therefore, we present an approach to detecting object regions considering horizontal and vertical edges. To this end, in the first step, image enhancement is performed by suppressing noises such as sea glint and complex clutters using a statistical filter. In the second step, a horizontal edge map and a vertical edge map are generated by 1-D Discrete Cosine Transform technique. Then, a combined map integrating the horizontal and the vertical edge maps is generated. In the third step, candidate object regions are detected by a adaptive thresholding method. Finally, exact object regions are extracted by eliminating background and clutter regions based on morphological operation.

• Journal of Korean Society for Geospatial Information Science
• /
• v.13 no.4 s.34
• /
• pp.3-9
• /
• 2005

The extraction of traffic information based on image processing is under broad research recently because the method based on image processing takes less cost and effort than the traditional method based on physical equipment. The main purpose of the algorithm based on image processing is to extract vehicles from an image correctly. Before the extraction, the algorithm needs the pre-processing such as background subtraction and binary image thresholding. During the pre-processing much noise is brought about because roadside tree and passengers in the sidewalk as well as vehicles are extracted as traffic flow. The noise undermines the overall accuracy of the algorithm. In this research, most of the noise could be removed by extracting the exact road area which does not include sidewalk or roadside tree. To extract the exact road area, traffic lanes in the image were used. Algorithm speed also increased. In addition, with the ratio between the sequential images, the problem caused by vehicles' shadow was minimized.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.48 no.4
• /
• pp.76-85
• /
• 2011

In this paper, we propose a texture feature-based language identification using Gabor feature and wavelet-domain BDIP (block difference of inverse probabilities) and BVLC (block variance of local correlation coefficients) features. In the proposed method, Gabor and wavelet transforms are first applied to a test image. The wavelet subbands are next denoised by Donoho's soft-thresholding. The magnitude operator is then applied to the Gabor image and the BDIP and BVLC operators to the wavelet subbands. Moments for Gabor magnitude image and each subband of BDIP and BVLC are computed and fused into a feature vector. In classification, the WPCA (whitened principal component analysis) classifier, which is usually adopted in the face identification, searches the training feature vector most similar to the test feature vector. Experimental results show that the proposed method yields excellent language identification with rather low feature dimension for a document image DB.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.38 no.6
• /
• pp.684-691
• /
• 2001

In this paper, we design and implement a system which recognizes and tracks a human head on a sequence of images. In this paper, the color of the skin and ellipse modeling is used as feature vectors to recognize the human head. And the modified time-varying edge detection method and the vertical projection method is used to acquire regions of the motion from images with very complex backgrounds. To select the head from the acquired candidate regions, the process for thresholding on the basis of the I-component of YIQ color information and mapping with ellipse modeling is used. The designed system shows an excellent performance in the cases of the rotated heads, occluded heads, and tilted heads as well as in the case of the normal up-right heads. And in this paper, the combinational technique of motion-based tracking and recognition-based tracking is used to track the human head exactly even though the human head moves fast.

• Journal of the Korean Society for Precision Engineering
• /
• v.18 no.11
• /
• pp.51-57
• /
• 2001

A light beam reflected from a machined surface generally containes information concerning about its surface roughness. This study examines and proposes a surface roughness measurement technique for on-machine measurement of machined surfaces. The technique is based on the measurement of a reflected laser beam pattern and the statistical analysis of its light intensity distribution. The surface roughness was found to be closely related to the standard deviation of the light intensity on the primary axis of the reflected pattern. An image acquisition device is made up of a laser diode, a half mirror, a screen, and a CCD camera. The exact image with the primary and secondary axes of a reflected laser beam pattern is calculated through such image processing algorithm as thresholding, edge detection, image rotation, segmentation, etc. A median filter and a surrounding light correction algorithm are improve the image quality and reduce the measuring error. Using the developed measuring device the effect of screen materials and workpiece and workpiece materials was investigated. Experimental results regarding to relatively high-quality surfaces machined by grinding, polishing, lapping processes have shown the measurement error is within 10% in the range of $0.1{mu}m~0.8{\mu}m R_q.$Therefore, the proposed method is thought to be effectively used when quick measurements is needed with workpieces fixed on the machine.

• Journal of the Institute of Convergence Signal Processing
• /
• v.5 no.4
• /
• pp.249-255
• /
• 2004

In fingerprint verification and identification, fingerprint and background region should be segmented. For this purpose, most systems obtain variance of brightness of X and Y direction using Sobel mask. To decide given local region is background or not, the variance is compared with a certain threshold. Although this method is simple, most fingerprint image does not separated with two region of fingerprint and background region. In this paper, we presented a new segmentation algorithm based on Run-Length Connectivity analysis. For a given binary image after thresholding, suggested algorithm calculates RL of X and Y direction. Until the given image is segmented to two regions, small run region is successively inverted. Experimental result show that this algorithm effectively separates fingerprint region and background region.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.24 no.1
• /
• pp.83-88
• /
• 2024

In this research, we propose a Semi-Supervised learning based railroad surface defect detection method. The Resnet50 model, pretrained on ImageNet, was employed for the training. Data without labels are randomly selected, and then labeled to train the ResNet50 model. The trained model is used to predict the results of the remaining unlabeled training data. The predicted values exceeding a certain threshold are selected, sorted in descending order, and added to the training data. Pseudo-labeling is performed based on the class with the highest probability during this process. An experiment was conducted to assess the overall class classification performance based on the initial number of labeled data. The results showed an accuracy of 98% at best with less than 10% labeled training data compared to the overall training data.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.12 no.4
• /
• pp.181-188
• /
• 2012

An effective extraction method for extracting a coated tongue is proposed in this paper, which is used as the diagnostic criteria in the tongue diagnosis. Proposed method uses the fluorescence response characteristics of the coated tongue that is occurred by using the ultraviolet light. Specially, this method can solved the previous problems including the issue in the limits of the diagnosis environment and in the objectivity of the diagnosis results. In our method, original tongue image is acquired by using the ultraviolet light, and binarization is performed by thresholding a valley-points in the histogram that corresponds to the color difference of tongue body and tongue coating. Final view image is presented to the oriental doctor, after applying the canny-edge algorithm to the binary image, and edge image is added to the original image. In order to evaluate the performance of the our proposed method, after building a various tongue image, we compared the true region of coated tongue by the oriental doctor's hand with the extracted region by the our method. As a result, the proposed method showed the average 87.87% extraction ratio. The shape of the extracted coated tongue region showed also significantly higher similarity.

• Journal of Broadcast Engineering
• /
• v.14 no.6
• /
• pp.749-768
• /
• 2009

In this paper, an intra prediction offset compensation method is proposed to improve intra prediction in H.264/AVC. In H.264/AVC, intra prediction based on various directions improves the coding efficiency by removing spatial correlation between neighboring blocks. In details, neighboring pixels in reconstructed block can be used as intra reference block for the current block to be coded when intra prediction method is used. In order to reduce further the prediction error of the intra reference block, the proposed method introduces an intra prediction offset which is determined in the sense of the rate-distortion optimization and is added to the conventional intra prediction block. Besides the intra prediction offset compensation, the coefficient thresholding method which is used for inter coding in JM 11.0, is used for chroma component in intra block, which leads the improvement of the luma coding efficiency of the proposed method. In experiments, we show that the proposed method achieves average 2.45% in High Profile condition and maximum 4.41% of bitrate reduction relative to JM 11.0.