• Journal of Korea Multimedia Society
• /
• v.22 no.1
• /
• pp.1-9
• /
• 2019

Single image rain removal is a typical inverse problem which decomposes the image into a background scene and a rain streak. Recent works have witnessed a substantial progress on the task due to the development of convolutional neural network (CNN). However, existing CNN-based approaches train the network with synthetically generated training examples. These data tend to make the network bias to the synthetic scenes. In this paper, we present an unsupervised framework for removing rain streaks from real-world rainy images. We focus on the natural phenomena that static rainy scenes capture a common background but different rain streak. From this observation, we train siamese network with the real rain image pairs, which outputs identical backgrounds from the pairs. To train our network, a real rainy dataset is constructed via web-crawling. We show that our unsupervised framework outperforms the recent CNN-based approaches, which are trained by supervised manner. Experimental results demonstrate that the effectiveness of our framework on both synthetic and real-world datasets, showing improved performance over previous approaches.

• Journal of Korea Multimedia Society
• /
• v.22 no.1
• /
• pp.99-109
• /
• 2019

Realistic rendering of natural phenomena is a difficult problem. Many environmental factors must be considered to simulate this phenomenon. At the same time, we need to think about their computational complexity to be simulated with computer algorithm One of the most difficult problems in creating weather conditions is the rain. To simulate realistic rainy scene, you have to consider the physical properties of rain and the environmental where the rain is falling down as well. In this paper, we survey the modeling and rendering techniques for realistic rainfall scenes from three different aspects. First, we list up techniques for modeling raindrop dynamics. Second, we survey the rendering techniques that render the raindrop in the environment. Third, we take a look at the hybrid methods that combines the rendering the modeling at the same time. For each aspect, we compare the algorithms in terms of implementation and their speciality.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.113-117
• /
• 2020

Images were taken under various weather such as rain, haze, snow often show low visibility, which can dramatically decrease accuracy of some tasks in computer vision: object detection, segmentation. Besides, previous work to enhance image usually downsample the image to receive consistency features but have not yet good upsample algorithm to recover original size. So, in this research, we jointly implement removal streak in heavy rain image and super resolution using a deep network. We put forth a 2-stage network: a multi-model network followed by a refinement network. The first stage using rain formula in the single image and two operation layers (addition, multiplication) removes rain streak and noise to get clean image in low resolution. The second stage uses refinement network to recover damaged background information as well as upsample, and receive high resolution image. Our method improves visual quality image, gains accuracy in human action recognition task in datasets. Extensive experiments show that our network outperforms the state of the art (SoTA) methods.

• Journal of Broadcast Engineering
• /
• v.20 no.1
• /
• pp.153-163
• /
• 2015

This paper proposes a rain detection and removal algorithm that is robust against camera motion in video sequences. In detection part, the proposed algorithm initially detects possible rain streaks by using intensity properties and spatial properties. Then, the rain streak candidates are selected based on Gaussian distribution model. In removal part, a non-rain block matching algorithm is performed between adjacent frames to find similar blocks to the block that has rain pixels. If the similar blocks to the block are obtained, the rain region of the block is reconstructed by non-local means (NLM) filter using the similar neighbors. Experimental results show that the proposed algorithm outperforms the previous works in terms of subjective visual quality of de-rained video sequences.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2013.06a
• /
• pp.12-13
• /
• 2013

본 논문은 비가 오는 장면을 촬영한 동영상에서 빗줄기를 효과적으로 제거하는 기법을 제안한다. 제안하는 기법에서는 광흐름 검출 기법을 이용하여 인접한 프레임에서 현재 프레임의 픽셀에 대응하는 픽셀을 검출하고, 확률에 따라 해당 프레임의 픽셀 적용 유무를 결정한다. 빗줄기로 검출된 픽셀을 인접한 프레임의 픽셀 값으로 대체함으로써 영상 내 빗줄기를 제거한다. 컴퓨터 모의실험을 통해 제안하는 알고리즘이 동영상에서 효과적으로 빗줄기를 제거할 수 있음을 확인한다.

• Atmosphere
• /
• v.19 no.1
• /
• pp.37-51
• /
• 2009

Kinematic characteristics of a heavy rainfall event occurred in Changma front are analyzed using synoptic weather charts, satellite imagery and NCEP(National Centers for Environmental Prediction) / NCAR(National Centers for Atmospheric Research) reanalysis data. The heavy rainfall is accompanied with mesoscale rain clouds developing over the Southwest region of Korea during the period from 0300 LST to 2100 LST 25 June 2006. The surface cyclone in the Changma front is generated and developed rapidly when it meets following vertical conditions: The maximum value of relative vorticity is appeared at 700 hPa and is extended gradually near the surface. It is thought that the vertical structure of relative vorticity is closely related with the descent of strong wind zone exceeding $10ms^{-1}$. The jet core at 200 hPa is shifted southward and extended downward and the low-level jet stream associated with upper-level jet stream appeared at 850 hPa. Kinematic features of heavy rainfall system at cyclone-generating point are as follows: In the generating stage of cyclone, the relative vorticity below 850 hPa increased and the convergence below 850 hPa and the divergence at 400 hPa are intensified by southward movement of jet core at 200 hPa. The heavy rainfall system seems to locate to the south of the exit region of upper-level jet streak; In the developing stage of cyclone, the relative vorticity below 850 hPa and the convergence near surface are further strengthened and upward vertical velocity between 850 hPa and 200 hPa is increased.

• Journal of Environmental Science International
• /
• v.21 no.12
• /
• pp.1511-1521
• /
• 2012

An attempt is made to analyse characteristic features of heavy rainfalls which occur at the metropolitan area of the Korean peninsular the on- and off- Changma season. For this, two representative heavy rainfall episodes are selected; one is the on-Changma season wherein a torrential rain episode happened at Goyang city on 12 July 2006, and the other is the off-Changma season, a heavy rainfall event in Seoul on 21 September 2006. Both recorded considerable amounts of precipitation, over 250mm in a half-day, which greatly exceeded the amount expected by numerical prediction models at those times, and caused great damage to property and life in the affected area. Similarities in the characteristics of both episodes were shown by; the location of upper-level jet streak and divergence fields of the upper wind over heavy rainfall areas, significantly high equivalent potential temperatures in the low atmospheric layer due to the entrainment of hot and humid air by the low-level jet, and the existence of very dry air and cold air pool in the middle layer of the atmosphere at the peak time of the rainfall events. Among them, differences in dynamic features of the low-level jet and the position of rainfall area along the low-level jet are remarkable.

• Journal of Environmental Science International
• /
• v.11 no.2
• /
• pp.111-120
• /
• 2002

Asian dust(or yellow sand) occurs mainly in spring and occasionally in winter in east Asia, when the weather conditions are under an upper trough/cut-off low and surface high/low pressure system during blocking episode days associated with the stationary patterns of the upper level jet stream. The transport mechanism for Asian dust during the blocking episode days in spring 2001 was analyzed using the TOMS aerosol index and meteorological mesoscale model 5(MM5). Based on the E vector, an extension of an Eliassen-Palm flux, the blocking episode days were found to be associated with the development of an upper cut-off low and surface cyclones. Concurrently, the occurrence of dust storms was also determined by strong cold advection at the rear of a jet streak, which exhibited a maximum wind speed within the upper jet stream. As such, the transport mechanism for Asian dust from China was due to advection of the isentropic potential vorticity(IPV) and isentropic surfaces associated with tropopause folding. The transport heights for Asian dust during the blocking episode days were found to be associated with the distribution of the isentropes below the IPV At the same time, lee waves propagated by topography affected the downward motion and blocking of Asian dust in China. The Asian dust transported from the dust source regions was deposited by fallout and rain-out with a reinforcing frontogenesis within a surface cyclone, as determined from satellite images using TOMS and GMS5. Accordingly, these results emphasize the importance of forecasting jet streaks, the IPV, and isentropes with geopotential heights in east Asia.