• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.205-207
• /
• 2020

This paper proposes a novel deep learning-based method to upsample a depth map. Most conventional methods estimate high-resolution depth map by modifying pixel value of given depth map using high-resolution color image and low-resolution depth map. However, these methods cause under- or over-shooting problems that restrict performance improvement. To overcome these problems, the proposed method iteratively performs grid warping scheme which shifts pixel values to restore blurred image for estimating high-resolution depth map. Experimental results show that the proposed method improves both quantitative and visual quality compared to the existing method.

• Journal of Broadcast Engineering
• /
• v.20 no.1
• /
• pp.40-47
• /
• 2015

This paper discusses QoS enhancement in the Best-effort services of the service plan provided by MPEG Media Transport (MMT) systems for video streaming applications. Among MMT services, i.e. per-flow, per-class, and best-effort services, the server does not provide guaranteed bandwidth for the best-effort service only. Therefore, in the best-effort services, a bandwidth access priority is defined for various services, where the lowest priority is assigned to the low-level video services. To alleviate the issue of bandwidth limitation in the best-effort services, this paper investigates transmission of low-resolution video with low bitrate and up-sampling. Our experimental results prove the superiority of the proposed method in terms of delivered video quality.

• Journal of Broadcast Engineering
• /
• v.25 no.2
• /
• pp.242-251
• /
• 2020

Recently, many deep convolutional neural networks for image super-resolution have been studied. Existing deep learning-based super-resolution algorithms are architecture that up-samples the resolution at the end of the network. The post-upsampling architecture has an inefficient structure at large scaling factor result of predicting a lot of information for mapping from low-resolution to high-resolution at once. In this paper, we propose a single image super-resolution using Channel Attention Residual Dense Block based on an iterative up-down sampling architecture. The proposed algorithm efficiently predicts the mapping relationship between low-resolution and high-resolution, and shows up to 0.14dB performance improvement and enhanced subjective image quality compared to the existing algorithm at large scaling factor result.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.55 no.7
• /
• pp.335-343
• /
• 2006

We have designed the system that measure one channel ECG by two electrode and extract real-time EDR with more related resipiration and comportable to subject by using conductive textile. On the assumption that relation between RL electrode and potential measurement electrode is coupled with RC connected model, we designed RL drive output to feedback two electrode for reduction of common mode signal. The conductive textile which was used for two ECG electrode was offered more comfort during night sleep in bed than any other method using attachments. In the method of single-lead EDR, R wave point or QRS interval area could be used for EDR estimation in traditional method, it is, so to speak, the amplitude modulation(AM) method for EDR. Alternatively, R-R interval could be used for frequency modulation(FM) method based on Respiratory Sinus Arrhythmia(RSA). For evaluation of performance on AM EDR and FM EDR from 14 subject, ECG lead III was measured. Each EDR was compared with both temperature around nose(direct measurement of respiration) and respiration signal from thoracic belt(indirect measurement of respiration) on mean squared error(MSE), cross correlation(Xcorr), and Coherence. The upsampling interpolation technique of multirate signal processing is applied to interpolating data instead of cubic spline interpolation. As a result, we showed the real-time EDR extraction processing to be implemented at micro-controller.

• Journal of Broadcast Engineering
• /
• v.20 no.3
• /
• pp.414-420
• /
• 2015

The quality of depth images is important in the 3D video system to represent complete 3D contents. However, the original depth image from a depth camera has a low resolution and a flickering problem which shows vibrating depth values in terms of temporal meaning. This problem causes an uncomfortable feeling when we look 3D contents. In order to solve a low resolution problem, we employ 3D warping and a depth weighted joint bilateral filter. A temporal mean filter can be applied to solve the flickering problem while we encounter a residual spectrum problem in the depth image. Thus, after classifying foreground andbackground regions, we use an upsampled depth image for a foreground region and temporal mean image for background region.Test results shows that the proposed method generates a time consistent depth video with a high resolution.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2012.07a
• /
• pp.40-43
• /
• 2012

3차원 영상 기술은 방송, 영화, 게임, 의료, 국방 등 다양한 기존 산업들과 융합하며 새로운 패러다임을 형성하고 있으며, 고품질 및 고해상도의 3차원 영상 획득에 대한 필요성이 강조되고 있다. 이에 따라, 최근에는 3차원 입체 영상을 제작 하는 방법 중 하나인 2D-plus-Depth 구조에 대한 연구가 활발히 진행되고 있다. 2D-plus-Depth 구조는 Charge-Coupled Device(CCD) 센서 등을 이용한 일반 카메라와 깊이 카메라를 결합한 형태로써 이 구조로부터 얻은 깊이 영상의 해상도를 상향 변환하기 위해서 Joint Bilateral Upsampling(JBU)[1], 컬러 영상의 정보를 활용한 보간법[2] 등의 방법들이 사용된다. 하지만 이 방법들은 깊이 영상을 높은 배율로 상향 변환할 경우 텍스처가 복사되거나 흐림 및 블록화 현상이 발생하는 문제점이 있다. 본 논문에서는 2D-plus-Depth 구조에서 얻은 고해상도 컬러 영상에서 보간 정보를 구하고 이 정보를 저해상도의 깊이 영상에 적용하여 상향 변환된 가이드 깊이 영상을 제작한다. 이 가이드 깊이 영상을 Bilateral Filtering[8]을 이용함으로써 고품질의 고해상도 깊이 영상을 획득한다. 실험 결과 제안하는 방법으로 해상도를 상향 변환을 할 경우에 기존의 보간법들에 비해 깊이 영상의 특성을 잘 보존함을 확인할 수 있고, 가이드 깊이 영상에 필터링을 처리한 결과가 JBU의 결과보다 향상됨을 확인할 수 있다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.29 no.10C
• /
• pp.1484-1489
• /
• 2004

Given a video frame in terms of its 8${\times}$8 block-DCT coefncients, we wish to obtain a downsized or upsized version of this Dame also in terms of 8${\times}$8 block DCT coefficients. The DCT being a linear unitary transform is distributive over matrix multiplication. This fact has been used for downsampling video frames in the DCT domains in Dugad's, Mukherjee's, and Park's methods. The downsampling and upsampling schemes combined together preserve all the low-frequency DCT coefficients of the original image. This implies tremendous savings for coding the difference between the original frame (unsampled image) and its prediction (the upsampled image).This is desirable for many applications based on scalable encoding of video. In this paper, we extend the earlier works to various DCT sizes, when we downsample and then upsample of an image by a factor of two. Through experiment, we could improve the PSM values whenever we increase the DCT block size. However, because the complexity will be also increase, we can say there is a tradeoff. The experiment result would provide important data for developing fast algorithms of compressed-domain image/video resizing.

• Korean Journal of Materials Research
• /
• v.32 no.8
• /
• pp.345-353
• /
• 2022

In this study, using deep learning, super-resolution images of transmission electron microscope (TEM) images were generated for nanomaterial analysis. 1169 paired images with 256 × 256 pixels (high resolution: HR) from TEM measurements and 32 × 32 pixels (low resolution: LR) produced using the python module openCV were trained with deep learning models. The TEM images were related to DyVO₄ nanomaterials synthesized by hydrothermal methods. Mean-absolute-error (MAE), peak-signal-to-noise-ratio (PSNR), and structural similarity (SSIM) were used as metrics to evaluate the performance of the models. First, a super-resolution image (SR) was obtained using the traditional interpolation method used in computer vision. In the SR image at low magnification, the shape of the nanomaterial improved. However, the SR images at medium and high magnification failed to show the characteristics of the lattice of the nanomaterials. Second, to obtain a SR image, the deep learning model includes a residual network which reduces the loss of spatial information in the convolutional process of obtaining a feature map. In the process of optimizing the deep learning model, it was confirmed that the performance of the model improved as the number of data increased. In addition, by optimizing the deep learning model using the loss function, including MAE and SSIM at the same time, improved results of the nanomaterial lattice in SR images were achieved at medium and high magnifications. The final proposed deep learning model used four residual blocks to obtain the characteristic map of the low-resolution image, and the super-resolution image was completed using Upsampling2D and the residual block three times.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.8
• /
• pp.2068-2082
• /
• 2023

With the rapid development of deep learning, Depth Map Super-Resolution (DMSR) method has achieved more advanced performances. However, when the upsampling rate is very large, it is difficult to capture the structural consistency between color features and depth features by these DMSR methods. Therefore, we propose a color-image guided DMSR method based on iterative depth feature enhancement. Considering the feature difference between high-quality color features and low-quality depth features, we propose to decompose the depth features into High-Frequency (HF) and Low-Frequency (LF) components. Due to structural homogeneity of depth HF components and HF color features, only HF color features are used to enhance the depth HF features without using the LF color features. Before the HF and LF depth feature decomposition, the LF component of the previous depth decomposition and the updated HF component are combined together. After decomposing and reorganizing recursively-updated features, we combine all the depth LF features with the final updated depth HF features to obtain the enhanced-depth features. Next, the enhanced-depth features are input into the multistage depth map fusion reconstruction block, in which the cross enhancement module is introduced into the reconstruction block to fully mine the spatial correlation of depth map by interleaving various features between different convolution groups. Experimental results can show that the two objective assessments of root mean square error and mean absolute deviation of the proposed method are superior to those of many latest DMSR methods.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.19 no.1
• /
• pp.660-665
• /
• 2018

Research into underwater wireless networks that enable the monitoring and controlling of the ocean environments has been continuing for disaster prevention and military proposes, as well as for the exploitation of ocean resources throughout the world. A research group led by Hoseo university has been studying a distributed underwater monitoring and controlling network. In this study, we developed an interpolator for acoustic communication between an underwater base station controller and underwater base station, which is included in this network. The underwater acoustic communication provided by this network defines four links whose sampling rates are different. Low power consumption is one of the most important requirements. Therefore, we adopted CIC interpolators, which are known to act as filters with a low power consumption, and some CIC interpolators with an appropriate changing rate were selected depending on the link. However, these interpolators have a large passband drop and wide transition region. To solve these problems, we added a compensator and half-band filter. After verifying the algorithm by using Matlab, we designed and verified it with Verilog-HDL in a ModelSim environment.