• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.2
• /
• pp.156-160
• /
• 2014

Thresholding in frequency domain is a simple and effective noise reduction technique. Determination of the threshold is critical to the image quality. The optimal threshold minimizing the Mean Square Error (MSE) is chosen adaptively in the wavelet domain; we utilize an equation of the MSE for the soft-thresholded signal and the histogram of wavelet coefficients of the original image and noisy image. The histogram of the original signal is estimated through the deconvolution assuming that the probability density functions (pdfs) of the original signal and the noise are statistically independent. The proposed method is quite general in that it does not assume any prior for the source pdf.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.50 no.6
• /
• pp.276-286
• /
• 2013

SVD-based deconvolution algorithm has been known as the most effective technique for CT perfusion image analysis. In this algorithm, in order to reduce noise effects, SVD coefficients smaller than a certain threshold are removed. As the truncation threshold, either a fixed value or a variable threshold yielding a predetermined OI (oscillation index) is frequently employed. Each of these two thresholding methods has an advantage to the other either in accuracy or efficiency. In this paper, we propose a Monte Carlo simulation method to evaluate the accuracy of the two methods. An extension of the proposed method is presented as well to measure the effects of image smoothing on the accuracy of the thresholding methods. In this paper, after the simulation method is described, experimental results are presented using both simulated data and real CT images.

• Proceedings of the Acoustical Society of Korea Conference
• /
• spring
• /
• pp.403-408
• /
• 1999

In this paper, ultrasonic tomography by the Mode-Converted Rayleigh wave (MCRW) in the back-scattered direction is presented. When a beam with a short pulse and narrow beam width enters a reflector with smooth surface, in general, two major arrivals can be observed in the output waveform: the specular reflection and the radiation of the MCRW from the reflector surface. The time-delay between the two waves is relatively large and thus can be measured easily. This large time-delay is due to the fact that the MCRW is slower than incident wave. In our method, this large time- delay is used for ultrasonic image reconstruction. To effectively detect the MCRW, the arrayed-receiving transducers are circularly arranged around the transmitter. In addition, a deconvolution method is employed to remove specular echo signals for reconstructing the MCRW image.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.5
• /
• pp.2253-2272
• /
• 2018

We propose a deep learning method for multi-focus image fusion. Unlike most existing pixel-level fusion methods, either in spatial domain or in transform domain, our method directly learns an end-to-end fully convolutional two-stream network. The framework maps a pair of different focus images to a clean version, with a chain of convolutional layers, fusion layer and deconvolutional layers. Our deep fusion model has advantages of efficiency and robustness, yet demonstrates state-of-art fusion quality. We explore different parameter settings to achieve trade-offs between performance and speed. Moreover, the experiment results on our training dataset show that our network can achieve good performance with subjective visual perception and objective assessment metrics.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.50 no.4
• /
• pp.127-136
• /
• 2013

This paper proposes a video deblurring algorithm which maintains uniform sharpness between frames. Unlike the previous algorithms using fixed parameters, the proposed algorithm keeps uniform sharpness by adjusting parameters for each frame. First, we estimate the initial blur kernel and perform deconvolution, then measure the sharpness of the deblurred image. In order to maintain uniform sharpness, we adjust the regularization parameter and kernel according to the examined sharpness, and perform deconvolution again. The experimental results show that the proposed algorithm achieves outstanding deblurring results while providing consistent sharpness.

• Journal of Astronomy and Space Sciences
• /
• v.39 no.4
• /
• pp.127-139
• /
• 2022

Irregular mare patches (IMPs), recently discovered on the Moon, are eruptions of magma on the lunar surface, and their origins are still in question. While prior studies on IMPs have mainly focused on optical image analysis, in this study, an analysis of the characteristics of minerals is performed exemplary for the first time. Modified Gaussian model (MGM) deconvolution was applied to the infrared spectrum to confirm the properties of the mafic mineral. Mafic minerals were analyzed for 6 olivine-rich (Ol-rich) IMPs out of 91 currently reported, and only 4 of them yielded results of significance. All four sites showed more abundance of Fe than Mg, and manifested a weak relationship with Mg-suite rock. However, a problem was discovered during the MGM application process due to pilot implementation. In order to solve this problem, it is required to adjust the MGM initial condition settings more precisely and to increase the signal to noise ratio of the observation data. Moreover, it is necessary to analyze the mineral properties for all IMPs considering minerals other than Ol and utilize them to deduce the origin of the IMPs.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.68-71
• /
• 2020

최근 딥러닝을 이용한 인체 자세 추정(human pose estimation) 연구가 활발히 진행되고 있다. 그 중 구조가 간단하면서도 성능이 강력하여 널리 사용되고 있는 딥러닝 네트워크 모델은 이미지 분류(image classification)에 사용되는 백본 네트워크(backbone network)와 디컨볼루션 출력망(deconvolution head network)을 이어 붙인 구조를 갖는다[1]. 기존의 디컨볼루션 출력망은 디컨볼루션 층을 쌓아 낮은 해상도의 특징맵을 모두 높은 해상도로 변환한 후 최종 인체 자세 추정을 하는데 이는 다양한 해상도에서 얻어낸 특징들을 골고루 활용하기 힘들다는 단점이 있다. 따라서 본 논문에서는 매 디컨볼루션 층 이후에 인체 자세 추정을 하여 다양한 해상도에서 연산을 하고 이를 종합하여 최종 인체 자세 추정을 하는 방법을 제안한다. 실험 결과 Res50 과 기존의 디컨볼루션 출력망의 경우 0.717 AP 를 얻었는데 Res101 과 기존의 디컨볼루션 출력망을 사용한 결과 50% 이상의 파라미터 수 증가와 함께 0.727 AP, 즉 0.010AP 의 성능 향상이 이루어졌다. 이에 반해 Res50 에 다중 해상도 디컨볼루션 출력망을 사용한 결과 약 1%의 파라미터 수 증가 만으로 0.720 AP, 즉 0.003 AP 의 성능 향상이 이루어졌다. 이를 통해 디컨볼루션 출력망 구조를 개선하면 매우 적은 파라미터 수 증가 만으로도 인체 자세 추정의 성능을 효과적으로 향상시킬 수 있음을 확인하였다.

• Journal of the Institute of Convergence Signal Processing
• /
• v.5 no.1
• /
• pp.18-24
• /
• 2004

SPECT technology is used for the reconstructed image in the field of industry noncontact measurement system. One of the distortion problems in reconstructed image quality is a collimator characterictic. The image distortion is caused by a geometrical structure of the collimator. This paper indicated a correction method to remove the image distortion by the structure of the collimator, and compared with the existing correction method. The correction. method removed the image distortion to use deconvolution of projection data with the shift-variant blurring function in the frequency domain. In this pater, I simulated with the collimator angle and distance between the detector and the center of object. and verified with expeimental data. The validity and limitation of correction method is studied for actual industrial applications.

• 한국HCI학회:학술대회논문집
• /
• 2009.02a
• /
• pp.27-33
• /
• 2009

We propose an efficient method for single image motion deblurring using edge prediction. Previous methods for motion deblurring from a single image have been based on total variation or natural image statistics. In contrast, our method predicts sharp edges by applying bilateral and shock filters and manipulating image gradients directly, and estimates motion blur using the predicted sharp edges. Sharp edge prediction makes our method possible to deblur efficiently with less computation. Results show that our method can effectively and efficiently restore images degraded by large complex motion blur.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.2
• /
• pp.855-870
• /
• 2019

In this paper, an adaptive iterative algorithm is proposed for motion deblurring by using the salient intensity prior. Based on the observation that the salient intensity of the clear image is sparse, and the salient intensity of the blurred image is less sparse during the image blurring process. The salient intensity prior is proposed to enforce the sparsity of the distribution of the saliency in the latent image, which guides the blind deblurring in various scenarios. Furthermore, an adaptive iteration strategy is proposed to adjust the number of iterations by evaluating the performance of the latent image and the similarity of the estimated blur kernel. The negative influence of overabundant iterations in each scale is effectively restrained in this way. Experiments on publicly available image deblurring datasets demonstrate that the proposed algorithm achieves state-of-the-art deblurring results with small computational costs.