• 한국정보디스플레이학회:학술대회논문집
• /
• 2007.08b
• /
• pp.1500-1503
• /
• 2007

Auto-stereoscopic 3D of 60-view number is made using slanted lenticular lens array and LCD of 15.1 inch diagonal size and 3200 by 2400 pixel numbers. Due to its large view number, smooth motion parallax is observed and the visual fatigue is reduced.

• Annual Conference of KIPS
• /
• 2016.10a
• /
• pp.735-736
• /
• 2016

본 논문에서는 웹캠을 이용하여 사용자가 추출하고자 하는 객체를 영상으로부터 추출하는 시스템을 제안한다. 영상의 중심에 위치한 Super-pixel을 표시 (Seed) 정보로 이용하여 영상에서 중심에 위치하고 있는 객체를 추출하는 방식을 제시한다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2019.06a
• /
• pp.255-256
• /
• 2019

본 논문에서는 초고해상도 기술의 복잡도를 줄이기 위하여 픽셀단위 생략 방법을 제안한다. 제안하는 방법은 픽셀 단위로 수평, 수직 방향의 밝기에 대한 2 차 미분치에 기반하여 생략을 결정한다. 제안하는 방법의 성능 평가를 위하여 가장 간단한 초고해상도 알고리즘인 SRCNN 과 제안하는 방법의 PSNR 비교한다. 그 결과 제안하는 방법이 평균적으로 약 47%의 픽셀이 생략이 되면서 SRCNN 대비 0.2dB PSNR 감소를 보인다.

• Journal of Broadcast Engineering
• /
• v.25 no.6
• /
• pp.954-964
• /
• 2020

In immersive media, the most important factor that provides immersion is depth information. Therefore, it is essential to obtain high quality depth information in order to produce high quality immersive media. In this paper we propose an algorithm to improve depth map, considering the segmentation of images and the relationship between multiple views in multi-view systems. The proposed algorithm uses a super-pixel segmentation technique to divide the depth map of the reference view into several segments, and project each segment into adjacent view. Subsequently, the depth map of the adjacent view is improved using plane estimation using the information of the projected segment, and then reversed to the reference view. This process is repeated for several adjacent views to improve the reference depth map by updating the values of the improved adjacent views and the initial depth map of the reference view. Through simulation, the proposed algorithm is shown to surpass the conventional algorithm subjectively and objectively.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.18 no.12
• /
• pp.2946-2952
• /
• 2014

Among the Example based Super Resolution(SR) techniques, Neighbor embedding(NE) has been inspired by manifold learning method, particularly locally linear embedding. However, the poor generalization of NE decreases the performance of such algorithm. The sizes of local training sets are always too small to improve the performance of NE. We propose the Learning Sparse-Neighbor Image Representation baesd on SVR having an excellent generalization ability to solve this problem. Given a low resolution image, we first use bicubic interpolation to synthesize its high resolution version. We extract the patches from this synthesized image and determine whether each patch corresponds to regions with high or low spatial frequencies. After the weight of each patch is obtained by our method, we used to learn separate SVR models. Finally, we update the pixel values using the previously learned SVRs. Through experimental results, we quantitatively and qualitatively confirm the improved results of the proposed algorithm when comparing with conventional interpolation methods and NE.

• Journal of Broadcast Engineering
• /
• v.9 no.2
• /
• pp.119-130
• /
• 2004

We have developed a prototype super-high-definition (SHD) digital cinema distribution system that can store, transmit, and display eight-million-pixel motion pictures that have the image quality of a 35-mm film movie. The system contains a movie server, a real-time decoder, and an SHB projector. Using a Gigabit Ethernet link and TCP/IP, the server transmits JPEG2000 compressed motion picture data streams to the decoder at transmission speeds as high as 300 Mbps. The received data streams are decompressed by the decoder, and then projected onto a screen via the projector. By using an enlarged TCP window, multiple TCP streams, and a shaping function to control the data transmission quantity, we achieved real-time streaming of SHD movie data at about 300 Mbps between Chicago and Los Angeles, a distance of more than 3000 km. We also improved the decoder performance to show movies with Image qualities of 450 Mbps or higher. Since UDP is more suitable than TCP for fast long-distance streaming, we have developed an SHD digital cinema UDP relay system, in which UDP is used for transmission over a fast long-distance network. By using four pairs of server-side-proxy and decoder-side-proxy, 450-Mbps movie data streams could be transmitted.

• The Bulletin of The Korean Astronomical Society
• /
• v.44 no.2
• /
• pp.70.4-70.4
• /
• 2019

Image super-resolution (SR) is a technique that enhances the resolution of a low resolution image. In this study, we use three SR models (RCAN, ProSRGAN and Bicubic) for enhancing solar SDO/HMI magnetograms using deep learning. Each model generates a high resolution HMI image from a low resolution HMI image (4 by 4 binning). The pixel resolution of HMI is about 0.504 arcsec. Deep learning networks try to find the hidden equation between low resolution image and high resolution image from given input and the corresponding output image. In this study, we trained three models with HMI images in 2014 and test them with HMI images in 2015. We find that the RCAN model achieves higher quality results than the other two methods in view of both visual aspects and metrics: 31.40 peak signal-to-noise ratio(PSNR), Correlation Coefficient (0.96), Root mean square error (RMSE) is 0.004. This result is also much better than the conventional bi-cubic interpolation. We apply this model to a full-resolution SDO/HMI image and compare the generated image with the corresponding Hinode NFI magnetogram. As a result, we get a very high correlation (0.92) between the generated SR magnetogram and the Hinode one.

• Journal of Digital Contents Society
• /
• v.15 no.6
• /
• pp.737-743
• /
• 2014

For single image super resolution (SR), interpolation based and example based algorithms are extensively used. The interpolation algorithms have the strength of theoretical simplicity. However, those algorithms are tending to produce high resolution images with jagged edges, because they are not able to use more priori information. Example based algorithms have been studied in the past few years. For example based SR, the nearest neighbor based algorithms are extensively considered. Among them, neighbor embedding (NE) has been inspired by manifold learning method, particularly locally linear embedding. However, the sizes of local training sets are always too small. So, NE algorithm is weak in the performance of the visuality and quantitative measure by the poor generalization of nearest neighbor estimation. An improved NE algorithm with Support Vector Regression (SVR) was proposed to solve this problem. Given a low resolution image, the pixel values in its high resolution version are estimated by the improved NE. Comparing with bicubic and NE, the improvements of 1.25 dB and 2.33 dB are achieved in PSNR. Experimental results show that proposed method is quantitatively and visually more effective than prior works using bicubic interpolation and NE.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.9
• /
• pp.1049-1059
• /
• 2019

We propose a learning-based image segmentation algorithm. Starting from super-pixels, our method learns the probability of merging two regions based on the ground truth made by humans. The learned information is used in determining whether the two regions should be merged or not in a segmentation stage. Unlike exiting learning-based algorithms, we use both local and object information. The local information represents features computed from super-pixels and the object information represent high level information available only in the learning process. The object information is considered as privileged information, and we can use a framework that utilize the privileged information such as SVM+. In experiments on the Berkeley Segmentation Dataset and Benchmark (BSDS 500) and PASCAL Visual Object Classes Challenge (VOC 2012) data set, out model exhibited the best performance with a relatively small training data set and also showed competitive results with a sufficiently large training data set.

• Korean Journal of Remote Sensing
• /
• v.13 no.1
• /
• pp.13-30
• /
• 1997

Recent improvements of satellite remote sensing sensors which are represented by hyperspectral imaging sensors and high spatial resolution sensors provide a large amount of data, typically several hundred megabytes per one scene. Moreover, increasing information exchange via internet and information super-highway requires the developments of more active service systems for processing and analysing of remote sensing data in order to provide value-added products. In this sense, an advanced satellite data processing system is being developed to achive high performance in computing speed and efficieney in processing a huge volume of data, and to make possible network computing and easy improving, upgrading and managing of systems. JAVA internet programming language provides several advantages for developing software such as object-oriented programming, multi-threading and robust memory managent. Using these features, a satellite data processing system named as GeoPixel has been developing using JAVA language. The GeoPixel adopted newly developed techniques including object-pipe connect method between each process and multi-threading structure. In other words, this system has characteristics such as independent operating platform and efficient data processing by handling a huge volume of remote sensing data with robustness. In the evaluation of data processing capability, the satisfactory results were shown in utilizing computer resources(CPU and Memory) and processing speeds.