• Journal of Korea Multimedia Society
• /
• v.22 no.10
• /
• pp.1149-1159
• /
• 2019

We present a novel deep learning architecture for obtaining a latent image from a single blurry image, which contains dynamic motion blurs through object/camera movements. The proposed architecture consists of two sub-modules: blur image restoration and optical flow estimation. The tasks are highly related in that object/camera movements make cause blurry artifacts, whereas they are estimated through optical flow. The ablation study demonstrates that training multi-task architecture simultaneously improves both tasks compared to handling them separately. Objective and subjective evaluations show that our method outperforms the state-of-the-arts deep learning based techniques.

• Journal of Sensor Science and Technology
• /
• v.28 no.3
• /
• pp.152-156
• /
• 2019

An ultraviolet (UV) image sensor is an extremely important optoelectronic device used in scientific and medical applications because it can detect images that cannot be obtained using visible or infrared image sensors. Because photodetectors and transistors are based on different materials, conventional UV imaging devices, which have a hybrid-type structure, require additional complex processes such as a backside etching of a GaN epi-wafer and a wafer-to-wafer bonding for the fabrication of the image sensors. In this study, we developed a monolithic GaN UV passive pixel sensor (PPS) by integrating a GaN-based Schottky-barrier type transistor and a GaN UV photodetector on a wafer. Both individual devices show good electrical and photoresponse characteristics, and the fabricated UV PPS was successfully operated under UV irradiation conditions with a high on/off extinction ratio of as high as $10^3$. This integration technique of a single pixel sensor will be a breakthrough for the development of GaN-based optoelectronic integrated circuits.

• Current Optics and Photonics
• /
• v.2 no.6
• /
• pp.547-553
• /
• 2018

In this paper, a new method for spatial filtering in digital holography is proposed and verified by simulations compared to conventional methods. The new method is based on the simultaneous acquisition of two digital holograms, which can be separated by distinct spatial modulation, in a single image. Two holograms are generated by two reference waves, which have different spatial modulation orientations. Then, the overlapping region between the DC term and the object wave in the first hologram can be replaced with a less-overlapping region of the object wave in the second hologram because the whole image contains two holograms where the same objective wave has been recorded. In the simulation results, it is confirmed that the reconstructed image by the new method has better quality than for the original method.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.2
• /
• pp.1064-1079
• /
• 2019

Information security has become increasingly important with the rapid development of mobile devices and internet. An efficient encryption system is a key to this end. In this paper, we propose an image encryption method based on the cross diffusion of two chaotic maps. We use two chaotic sequences, namely the Logistic map and the Chebyshev map, for key generation which has larger security key space than single one. Moreover, we use these two sequences for further image encryption diffusion which decreases the correlation of neighboring pixels significantly. We conduct extensive experiments on several well-known images like Lena, Baboon, Koala, etc. Experimental results show that our algorithm has the characteristics of large key space, fast, robust to statistic attack, etc.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2018.06a
• /
• pp.63-66
• /
• 2018

Recently, a convolutional neural network (CNN) models at single image super-resolution have been very successful. Residual learning improves training stability and network performance in CNN. In this paper, we compare four convolutional neural network models for super-resolution (SR) to learn nonlinear mapping from low-resolution (LR) input image to high-resolution (HR) target image. Four models include general CNN model, global residual learning CNN model, local residual learning CNN model, and the CNN model with global and local residual learning. Experiment results show that the results are greatly affected by how skip connections are connected at the basic CNN network, and network trained with only global residual learning generates highest performance among four models at objective and subjective evaluations.

• International Journal of Advanced Culture Technology
• /
• v.8 no.4
• /
• pp.229-234
• /
• 2020

Dehazing significantly improves image quality by restoring the loss of contrast and color saturation for images taken in the presence. However, when applied to images not taken according to the prior information, dehazing can cause unintended degradation of image quality. To avoid unintended degradations, we present a hazy scene detection algorithm using a single image based on the distributions of hue, saturation, and dark channel. Through a heuristic approach, we find out statistical characteristics of the distribution of hue, saturation, and dark channels in the hazy scene and make a detection model using them. The proposed method can precede the dehazing to prevent unintended degradation. The detection performance evaluated with a set of test images shows a high hit rate with a low false alarm ratio. Ultimately the proposed method can be used to control the effect of dehazing so that the dehazing can be applied to wide variety of images without unintended degradation of image quality.

• Progress in Medical Physics
• /
• v.32 no.1
• /
• pp.1-17
• /
• 2021

The evolution of X-ray computed tomography (CT) has been based on the discovery of X-rays, the inception of the Radon transform, and the development of X-ray digital data acquisition systems and computer technology. Unlike conventional X-ray imaging (general radiography), CT reconstructs cross-sectional anatomical images of the internal structures according to X-ray attenuation coefficients (approximate tissue density) for almost every region in the body. This article reviews the essential physical principles and technical aspects of the CT scanner, including several notable evolutions in CT technology that resulted in the emergence of helical, multidetector, cone beam, portable, dual-energy, and phase-contrast CT, in integrated imaging modalities, such as positron-emission-tomography-CT and single-photon-emission-computed-tomography-CT, and in clinical applications, including image acquisition parameters, CT angiography, image adjustment, versatile image visualizations, volumetric/surface rendering on a computer workstation, radiation treatment planning, and target localization in radiotherapy. The understanding of CT characteristics will provide more effective and accurate patient care in the fields of diagnostics and radiotherapy, and can lead to the improvement of image quality and the optimization of exposure doses.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• fall
• /
• pp.41-44
• /
• 2021

In this paper, we propose an efficient reference-based compression artifact reduction network for 360-degree images in an equi-rectangular projection (ERP) domain. In our insight, conventional image restoration methods cannot be applied straightforwardly to 360-degree images due to the spherical distortion. To address this problem, we propose an adaptive disparity estimator using a deformable convolution to exploit correlation among 360-degree images. With the help of the proposed convolution, the disparity estimator establishes the spatial correspondence successfully between the ERPs and extract matched textures to be used for image restoration. The experimental results demonstrate that the proposed algorithm provides reliable high-quality textures from the reference and improves the quality of the restored image as compared to the state-of-the-art single image restoration methods.

• International Journal of Internet, Broadcasting and Communication
• /
• v.15 no.4
• /
• pp.166-171
• /
• 2023

With significant advancements in the field of artificial intelligence, many novel algorithms and technologies have emerged. Currently, AI painting can generate high-quality images based on textual descriptions. However, it is often challenging to control details when generating images, even with complex textual inputs. Therefore, there is a need to implement additional control mechanisms beyond textual descriptions. Based on ControlNet, this passage describes a combined utilization of various local controls (such as edge maps and depth maps) and global control within a single model. It provides a comprehensive exposition of the fundamental concepts of ControlNet, elucidating its theoretical foundation and relevant technological features. Furthermore, combining methods and applications, understanding the technical characteristics involves analyzing distinct advantages and image differences. This further explores insights into the development of image generation patterns.

• Journal of the Semiconductor & Display Technology
• /
• v.22 no.3
• /
• pp.149-154
• /
• 2023

This paper presents a novel approach to multi-focus image fusion using light field cameras. The proposed neural network, LFFCNN (Light Field Focus Convolutional Neural Network), is composed of three main modules: feature extraction, feature fusion, and feature reconstruction. Specifically, the feature extraction module incorporates SPP (Spatial Pyramid Pooling) to effectively handle images of various scales. Experimental results demonstrate that the proposed model not only effectively fuses a single All-in-Focus image from images with multi focus images but also offers more efficient and robust focus fusion compared to existing methods.