• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.07a
• /
• pp.721-723
• /
• 2020

부분 공간 군집화는 고차원 데이터에서 의미 있는 특징들을 선별 및 추출하여 저차원의 부분 공간에서 군집화 하는 것이다. 그러나 최근 딥러닝 활용한 부분 공간 군집화 연구들은 AutoEncoder을 기반으로 의미있는 특징을 선별하는 것이 아닌 특징 맵의 크기를 증가시켜서 네트워크의 표현 능력에 중점을 둔 연구되고 있다. 본 논문에서는 AutoEncdoer 네트워크에 Channel Attention 모델을 활용하여 Encoder와 Decoder에서 부분 공간 군집화를 위한 특징을 강조하는 네트워크를 제안한다. 본 논문에서 제안하는 네트워크는 고차원의 이미지에서 부분 공간 군집화를 위해 강조된 특징 맵을 추출하고 이를 이용해서 보다 향상된 성능을 보여주었다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.9
• /
• pp.2626-2641
• /
• 2024

Images taken in low lamination condition suffer from low contrast and loss of information. Low lumination image enhancement algorithms are required to improve the quality and broaden the applications of such images. In this study, we proposed a new Low lumination image enhancement architecture consisting of a transformer-based curve learning and an encoder-decoder-based texture enhancer. Considering the high effectiveness of curve matching, we constructed a transformer-based network to estimate the learnable curve for pixel mapping. Curve estimation requires global relationships that can be extracted through the transformer framework. To further improve the texture detail, we introduced an encoder-decoder network to extract local features and suppress the noise. Experiments on LOL and SID datasets showed that the proposed method not only has competitive performance compared to state-of-the-art techniques but also has great efficiency.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.10
• /
• pp.4080-4097
• /
• 2020

In recent years, convolutional neural network (CNN) has become the primary method for face detection. But its shortcomings are obvious, such as expensive calculation, heavy model, etc. This makes CNN difficult to use on the mobile devices which have limited computing and storage capabilities. Therefore, the design of lightweight CNN for face detection is becoming more and more important with the popularity of smartphones and mobile Internet. Based on the CPU real-time face detector FaceBoxes, we propose a multi-task lightweight face detector, which has low computing cost and higher detection precision. First, to improve the detection capability, the squeeze and excitation modules are used to extract attention between channels. Then, the textual and semantic information are extracted by shallow networks and deep networks respectively to get rich features. Finally, the landmark detection module is used to improve the detection performance for small faces and provide landmark data for face alignment. Experiments on AFW, FDDB, PASCAL, and WIDER FACE datasets show that our algorithm has achieved significant improvement in the mean average precision. Especially, on the WIDER FACE hard validation set, our algorithm outperforms the mean average precision of FaceBoxes by 7.2%. For VGA-resolution images, the running speed of our algorithm can reach 23FPS on a CPU device.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.7
• /
• pp.1888-1906
• /
• 2024

Aiming at the problems that the edge of melanoma image is fuzzy, the contrast with the background is low, and the hair occlusion makes it difficult to segment accurately, this paper proposes a model MSCNet for melanoma segmentation based on U-net frame. Firstly, a multi-scale pyramid fusion module is designed to reconstruct the skip connection and transmit global information to the decoder. Secondly, the contextural information conduction module is innovatively added to the top of the encoder. The module provides different receptive fields for the segmented target by using the hole convolution with different expansion rates, so as to better fuse multi-scale contextural information. In addition, in order to suppress redundant information in the input image and pay more attention to melanoma feature information, global channel attention mechanism is introduced into the decoder. Finally, In order to solve the problem of lesion class imbalance, this paper uses a combined loss function. The algorithm of this paper is verified on ISIC 2017 and ISIC 2018 public datasets. The experimental results indicate that the proposed algorithm has better accuracy for melanoma segmentation compared with other CNN-based image segmentation algorithms.

• Current Optics and Photonics
• /
• v.8 no.5
• /
• pp.463-471
• /
• 2024

Multimode fibers (MMFs) possess high information throughput and small core diameter, making them highly promising for applications such as endoscopy and communication. However, modal dispersion hinders the direct use of MMFs for image transmission. By training neural networks on time-series waveforms collected from MMFs it is possible to reconstruct images, transforming blurred speckle patterns into recognizable images. This paper proposes a fully convolutional neural-network model, MSMDFNet, for image restoration in MMFs. The network employs an encoder-decoder architecture, integrating multiscale convolutional modules in the decoding layers to enhance the receptive field for feature extraction. Additionally, attention mechanisms are incorporated from both spatial and channel dimensions, to improve the network's feature-perception capabilities. The algorithm demonstrates excellent performance on MNIST and Fashion-MNIST datasets collected through MMFs, showing significant improvements in various metrics such as SSIM.

• Journal of IKEEE
• /
• v.15 no.1
• /
• pp.70-75
• /
• 2011

Longitudinal recording has been the cornerstone of all two generations of magnetic recording systems, FDD and HDD. In recent, perpendicular recording has received much attention as promising technology for future high-density recording system Research into signal processing techniques is paramount for the issued storage system and is indispensable like longitudinal recording systems. This paper focuses on the performance evaluation of the various detectors under perpendicular recording system. Parameters for improving the their performance are examined for some detectors. Detectors considered in this work are the partial response maximum likelihood (PRML), noise-predictive maximum likelihood (NPML), fixed delay tree search with decision feedback (FDTS/DF), dual decision feedback equalizer (DDFE) and multilevel decision feedback equalizer (MDFE). Their performances are analyzed in terms of mean squared error (MSE) and noise power spectra, and similarity between recording channel and partial response (PR) channel.

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.29 no.10 s.241
• /
• pp.1083-1091
• /
• 2005

Turbulent flow around a rotating circular cylinder is investigated by Direct Numerical Simulation. The calculation is performed at three cases of low Reynolds number, Re=161, 348 and 623, based on the cylinder radius and friction velocity. Statistically strong similarities with fully developed channel flow are observed. Instantaneous flow visualization reveals that the turbulence length scale typically decreases as Reynolds number increases. Some insight into the spacial characteristics in conjunction with wave number is provided by wavelet analysis. The budget of dissipation rate as well as turbulent kinetic energy is computed and particular attention is given to the comparison with plane channel flow.

• 한국정보통신설비학회:학술대회논문집
• /
• 2007.08a
• /
• pp.41-45
• /
• 2007

Multi-input multi-output (MIMO) system can increase data rate, capacity and bit error rate (BER) performance compare to traditional single antenna system. However MIMO technique is pointed out the problem that has high complexity to design receiver. So a recent trend of research on the MIMO system pays more attention to simplified implementation of receiver structure. In this paper, we propose differential space time block code (STBC) for MIMO system with cyclic delay diversity (CDD). This structure can provide a very close performance to that of the conventional diversity scheme with maximum likelihood (ML) detection without channel estimation block while the receiver structure is highly simplified. Bit error rate (BER) performance of the proposed system is simulated for an AWGN channel by theoretical and simulated approaches. The results of this paper can be applicable to the 4G mobile multimedia communication systems.

• Proceedings of the IEEK Conference
• /
• 2002.07b
• /
• pp.1316-1319
• /
• 2002

In this paper, an online SNR estimator is derived for turbo coded multicode DS/SS systems in Nakagami fading channels. The multicode DS/SS approach is one of promising solutions to obtain higher-rate data transmission in DS/SS technologies. Turbo coding has paid much attention because of the significant improvements on error rate performances in various communication systems including multicode DS/SS systems. However, in the turbo decoding, channel state information, especially signal-to-noise ratio (SNR) at the correlator outputs, is desired in order to obtain such improvements. We evaluate the accuracy of the derived SNR estimation. It is shown that the bit error rate performance using our SNR estimation is close to the performance with perfect knowledge of channel state information.

• Proceedings of the KSME Conference
• /
• 2002.08a
• /
• pp.143-146
• /
• 2002

Recently the micro flows have been received much attention in the applications with regard to Micro Electro Mechanical Systems(MEMS). Such flows are governed by relatively large viscous forces, compared with inetia force, and frequently specified by slip wall boundary conditions. In the present study, computations are applied to investigate the compressible viscous flows through an extremely small channel, and theoretical analyses are conducted using the Fanno flow theory. The axisymmetic, compressible, Wavier-Stokes equations are numerically solved by a fully implicit finite implicit method. The predicted results are validated with previous experimental data available. The results obtained show that for small Reynolds numbers, the viscous frictional force significantly influences the compressible micro channel flows.