• 한국정보디스플레이학회:학술대회논문집
• /
• 2005.07a
• /
• pp.211-216
• /
• 2005

High-density directional display provides natural three-dimensional images. A large number of directional images are displayed in different horizontal directions with directional rays. There are two different types of display configurations. One is the projection-type and the other is the thin-type. The 64-directional and 128-directional displays using the projection-type configuration and the 72-directional display using the thin-type configuration are presented. The human responses are also shown.

• Journal of Korea Multimedia Society
• /
• v.4 no.6
• /
• pp.534-542
• /
• 2001

Recent content-based image retrieval systems make use of a local single server contained a limited number of images. So these systems are not satisfactory for the Web user's needs that make request for various images on the Web. A content-based image retrieval system that has regard for a great number of Web images has to stand on the basis of real-time first of all. Therefore, to implement the above system we have to resolve a problem of large waste time to take for an image collection and feature extractions. In recent, PC clusters with a load distribution are implemented for the purpose of high-performance data processing. In this paper, we decreased the whole retrieval time by distributing the tasks of image collection and feature extraction to take much time among the slave computers of the PC cluster, and so we found the possibility of the real-time processing in the retrieval of Web images.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.12
• /
• pp.4326-4344
• /
• 2021

The SinGAN is one of generative adversarial networks that can be trained on a single nature image. It has poor ability to learn more global features from nature image, and losses much local detail information when it generates arbitrary size image sample. To solve the problem, a non-linear function is firstly proposed to control downsampling ratio that is ratio between the size of current image and the size of next downsampled image, to increase the ratio with increase of the number of downsampling. This makes the low-resolution images obtained by downsampling have higher proportion in all downsampled images. The low-resolution images usually contain much global information. Therefore, it can help the model to learn more global feature information from downsampled images. Secondly, the attention mechanism is introduced to the generative network to increase the weight of effective image information. This can make the network learn more local details. Besides, in order to make the output image more natural, the TVLoss function is introduced to the loss function of SinGAN, to reduce the difference between adjacent pixels and smear phenomenon for the output image. A large number of experimental results show that our proposed model has better performance than other methods in generating random samples with fixed size and arbitrary size, image harmonization and editing.

• Nuclear Engineering and Technology
• /
• v.53 no.5
• /
• pp.1626-1633
• /
• 2021

In an industrial field, non-destructive testing (NDT) is commonly used to inspect industrial products. Among NDT methods using radiation sources, X-ray laminography has several advantages, such as high depth resolution and low computational costs. Moreover, an X-ray laminography system with stationary source array and compact detector is able to reduce mechanical motion artifacts and improve inspection efficiency. However, this system, called stationary inverse-geometry X-ray laminography (s-IGXL), causes truncation artifacts in reconstructed images due to limited fields-of-view (FOVs). In this study, we proposed a projection data correction (PDC) method to reduce the truncation artifacts arisen in s-IGXL images, and the performance of the proposed method was evaluated with the different number of focal spots in terms of quantitative accuracy. Comparing with conventional techniques, the PDC method showed superior performance in reducing truncation artifacts and improved the quantitative accuracy of s-IGXL images for all the number of focal spots. In conclusion, the PDC method can improve the accuracy of s-IGXL images and allow precise NDT measurements.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.12 no.3
• /
• pp.250-255
• /
• 2012

Face recognition has wide applications in security and surveillance systems as well as in robot vision and machine interfaces. Conventional challenges in face recognition include pose, illumination, and expression, and face recognition at a distance involves additional challenges because long-distance images are often degraded due to poor focusing and motion blurring. This study investigates the effectiveness of applying photon-counting linear discriminant analysis (Pc-LDA) to face recognition in harsh environments. A related technique, Fisher linear discriminant analysis, has been found to be optimal, but it often suffers from the singularity problem because the number of available training images is generally much smaller than the number of pixels. Pc-LDA, on the other hand, realizes the Fisher criterion in high-dimensional space without any dimensionality reduction. Therefore, it provides more invariant solutions to image recognition under distortion and degradation. Two decision rules are employed: one is based on Euclidean distance; the other, on normalized correlation. In the experiments, the asymptotic equivalence of the photon-counting method to the Fisher method is verified with simulated data. Degraded facial images are employed to demonstrate the robustness of the photon-counting classifier in harsh environments. Four types of blurring point spread functions are applied to the test images in order to simulate long-distance acquisition. The results are compared with those of conventional Eigen face and Fisher face methods. The results indicate that Pc-LDA is better than conventional facial recognition techniques.

• Journal of information and communication convergence engineering
• /
• v.15 no.4
• /
• pp.256-261
• /
• 2017

In this paper, we propose a modified smart pixel mapping (SPM) to visualize occluded three-dimensional (3D) objects in real image fields. In integral imaging, orthoscopic real 3D images cannot be displayed because of lenslets and the converging light field from elemental images. Thus, pseudoscopic-to-orthoscopic conversion which rotates each elemental image by 180 degree, has been proposed so that the orthoscopic virtual 3D image can be displayed. However, the orthoscopic real 3D image cannot be displayed. Hence, a conventional SPM that recaptures elemental images for the orthoscopic real 3D image using virtual pinhole array has been reported. However, it has a critical limitation in that the number of pixels for each elemental image is equal to the number of elemental images. Therefore, in this paper, we propose a modified SPM that can solve this critical limitation in a conventional SPM and can also visualize the occluded objects efficiently.

• Journal of Korea Game Society
• /
• v.5 no.1
• /
• pp.53-60
• /
• 2005

To produce deep, entertaining mobile content, a large number of images must be included. But because of the limits on runtime memory for mobile phones, images cannot be used as easily in a mobile environment as they are in a computer. Therefore in this paper, I propose several different methods for efficiently using images in a mobile environment. The various methods I propose for using images are: Creating images using compression/decompression and rotation/symmetry Creating images of several different colors by changing the palette index of a bitmap Creating images through image combination Creating background images by using tile maps Creating new images through effects.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.35S no.2
• /
• pp.97-105
• /
• 1998

Subband/DCT coding has been introduced in order to transmit images of various resultions using one given image-codec, for nowadays there are various grades of quality in visual communication services. However, subband/DCT results in the increawse of multiplication number and memory size. In order to resolve this problem, we propose block-based subband/DCT coding in this paper. In block-based subband/DCT, the number of multiplications is not only reduced because we combine subband decomposistion with DCT, but the size of memory is also reduced because images can be parallel-processed block by block. We show that the number of multiplications is reduced, by analyzing the property ofblock-based subband/DCT matrix mathematically, and examine the performance of proposed coder, which adopts JPEG as backhand-coder after block-based subband/DCT.

• Progress in Medical Physics
• /
• v.16 no.4
• /
• pp.161-165
• /
• 2005

The CT number corresponds to electron density and its influence on dose calculation was studied. Five kinds of CT scanners were used to obtain Images of electron density calibration phantom (Gammex RMI 467), Then the differences between CT numbers for each scanners were ${\pm}2\%$ In homogeneous medium and $9.5\%$ in high density medium. In order to Investigate the influence of CT number to dose calculation, patients' thoracic CT images were analyzed. The maximum dose difference was $0.48\%$ for each organ. It acquired the phantom Images inserted high density material in the water phantom. Comparing the doses calculated with CT Images from each CT scanner, the maximum dose difference was $2.1\%$ in 20 cm in depth. The exact density to CT number conversion according to CT scanner is required to minimize the uncertainty of dose depends on CT number Especially the each hospital with various CT scanners has to discriminate CT numbers for each CT scanner. Moreover a periodic quality assurance is required for reproducibility of CT number.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.6
• /
• pp.1818-1832
• /
• 2022

Violence in the Internet era poses a new challenge to the current counter-riot work, and according to research and analysis, most of the violent incidents occurring are related to the dissemination of violence images. The use of the popular deep learning neural network to automatically analyze the massive amount of images on the Internet has become one of the important tools in the current counter-violence work. This paper focuses on the use of transfer learning techniques and the introduction of an attention mechanism to the residual network (ResNet) model for the classification and identification of violence images. Firstly, the feature elements of the violence images are identified and a targeted dataset is constructed; secondly, due to the small number of positive samples of violence images, pre-training and attention mechanisms are introduced to suggest improvements to the traditional residual network; finally, the improved model is trained and tested on the constructed dedicated dataset. The research results show that the improved network model can quickly and accurately identify violence images with an average accuracy rate of 92.20%, thus effectively reducing the cost of manual identification and providing decision support for combating rebel organization activities.