• Journal of Korea Multimedia Society
• /
• v.17 no.7
• /
• pp.838-848
• /
• 2014

In this paper, we conducted a cause analysis on boundary artifacts in image deconvolution. Results of the cause analysis show that boundary artifacts are caused not only by a misuse of boundary conditions but also by no use of the normalized backprojection. Results also showed that the correct use of boundary conditions does not necessarily remove boundary artifacts. Based on these observations, we suggest not to use any specific boundary conditions and to use the normalized backprojector for boundary artifact-free image deconvolution.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.12
• /
• pp.130-139
• /
• 2014

Undesired camera shaking can make a blur effect, which causes a degradation of video quality. We propose an efficient method of removing the blur effects on video captured from a single camera. The proposed method has a sequential process that is applied to each frame. The first stage is to estimate the camera motion for each frame. In order to estimate the camera motion, we compute the optical flow using 3 consecutive frames. Then a patch-wise image deconvolution is applied. During the deconvolution, edge prediction is used to improve the quality of image deconvolution. After patch-wise image deconvolution, deblurred patches are integrated into an image to produce a deblurred frame. The above process is performed for each frame. The experimental result shows that the proposed method removes the blur effect efficiently.

• Journal of Korea Multimedia Society
• /
• v.20 no.11
• /
• pp.1750-1758
• /
• 2017

In super-resolution, various methods with Convolutional Neural Network(CNN) have recently been proposed. CNN based methods provide much higher image quality than conventional methods. Especially, VDSR outperforms other CNN based methods in terms of image quality. However, it requires a high computational complexity which prevents real-time processing. In this paper, the method to apply a deconvolution layer to VDSR is proposed to reduce computational complexity. Compared to original VDSR, the proposed method achieves the 4.46 times speed-up and its degradation in image quality is less than -0.1 dB which is negligible.

• IEIE Transactions on Smart Processing and Computing
• /
• v.4 no.6
• /
• pp.413-421
• /
• 2015

Regularized Blind Deconvolution is a problem applicable in degraded images in order to bring the original image out of blur. Multichannel blind Deconvolution considered as an optimization problem. Each step in the optimization is considered as variable splitting problem using an algorithm called Alternating Minimization Algorithm. Each Step in the Variable splitting undergoes Augmented Lagrangian method (ALM) / Bregman Iterative method. Regularization is used where an ill posed problem converted into a well posed problem. Two well known regularizers are Tikhonov class and Total Variation (TV) / L2 model. TV can be isotropic and anisotropic, where isotropic for L2 norm and anisotropic for L1 norm. Based on many probabilistic model and Fourier Transforms Image deblurring can be solved. Here in this paper to improve the performance, we have used an adaptive regularization filtering and isotropic TV model Lp norm. Image deblurring is applicable in the areas such as medical image sensing, astrophotography, traffic signal monitoring, remote sensors, case investigation and even images that are taken using a digital camera / mobile cameras.

• Proceedings of the IEEK Conference
• /
• 2000.09a
• /
• pp.389-392
• /
• 2000

In this paper, we propose a novel post processing method of deconvolution for SAR images based on phase extension inverse filtering, which improves spatial resolution as well as effectively eliminates sidelobes with low computational complexity. It extends the bandwidth only to control the magnitude of the processed SAR data without distortions of the phase in frequency domain unlike the other techniques such as spatially variant apodization (SVA), and other deconvolution techniques. We compare the image processed by the proposed method with images processed by uniform weighting function, Hamming weighting function whose coefficient is 0.75, and SVA.

• Journal of applied mathematics & informatics
• /
• v.19 no.1_2
• /
• pp.151-164
• /
• 2005

The regularized structured total least norm (RSTLN) method finds an approximate solution x and error matrix E to the overdetermined linear system (H + E)x $\approx$ b, preserving structure of H. A new separation scheme by parts of variables for the regularized structured total least norm on blind deconvolution problem is suggested. A method combining the regularized structured total least norm method with a separation by parts of variables can be obtain a better approximated solution and a smaller residual. Computational results for the practical problem with Block Toeplitz with Toeplitz Block structure show the new method ensures more efficiency on image restoration.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.23 no.6
• /
• pp.609-617
• /
• 2020

Non-destructive inspection using X-rays is used as a method to check the inside of products. In order to accurately inspect, a X-ray image requires a higher spatial resolution. However, the reduction in pixel size of the X-ray detector, which determines the spatial resolution, is time-consuming and expensive. In this regard, a DMM has been proposed to obtain an improved spatial resolution using the same X-ray detector. However, this has a limitation that the motion blur phenomenon, which is a decrease in spatial resolution. In this paper, motion blur was removed by applying Non-Blind Deconvolution to the DMM image, and the increase in spatial resolution was confirmed. DMM and Non-Blind Deconvolution were sequentially applied to X-ray images, confirming 62 % MTF value by an additional 29 % over 33 % of DMM only. In addition, SSIM and PSNR were compared to confirm the similarity to the 1/2 pixel detector image through 0.68 and 33.21 dB, respectively.

• Journal of Korea Multimedia Society
• /
• v.21 no.12
• /
• pp.1363-1369
• /
• 2018

Breast cancer is one of the most common cancers in women worldwide. In Korea, breast cancer is most common cancer in women followed by thyroid cancer. The purpose of this study is to evaluate the possibility of using deep - run model for segmentation of breast masses and to identify the best deep-run model for breast mass segmentation. In this study, data of patients with breast masses were collected at Asan Medical Center. We used 596 images of mammography and 596 images of gold standard. In the area of interest of the medical image, it was cut into a rectangular shape with a margin of about 10% up and down, and then converted into an 8-bit image by adjusting the window width and level. Also, the size of the image was resampled to $150{\times}150$. In Deconvolution net, the average accuracy is 91.78%. In U-net, the average accuracy is 90.09%. Deconvolution net showed slightly better performance than U-net in this study, so it is expected that deconvolution net will be better for breast mass segmentation. However, because of few cases, there are a few images that are not accurately segmented. Therefore, more research is needed with various training data.

• Journal of Korea Multimedia Society
• /
• v.20 no.6
• /
• pp.865-872
• /
• 2017

Recent advance in electronics and scintillators makes it possible to utilize the time-of-flight (TOF) information in improving image reconstruction of positron emission tomography(PET). In this paper, we propose a TOF-based fast image reconstruction method for PET. The proposed method uses the deconvolution of TOF data for each angle view and the rotational averaging of deconvolved images. Simulation results show an improved performance of the proposed method, as compared with filtered backprojection (FBP) method, TOF-FBP, and TOF version of expectation-maximization(EM) methods. Simulation results also show a great potentiality of the proposed method in limited angle tomography applications.

• Journal of Information Processing Systems
• /
• v.19 no.1
• /
• pp.98-108
• /
• 2023

This paper is an attempt to design segmentation method based on fully convolutional networks (FCN) and attention mechanism. The first five layers of the Visual Geometry Group (VGG) 16 network serve as the coding part in the semantic segmentation network structure with the convolutional layer used to replace pooling to reduce loss of image feature extraction information. The up-sampling and deconvolution unit of the FCN is then used as the decoding part in the semantic segmentation network. In the deconvolution process, the skip structure is used to fuse different levels of information and the attention mechanism is incorporated to reduce accuracy loss. Finally, the segmentation results are obtained through pixel layer classification. The results show that our method outperforms the comparison methods in mean pixel accuracy (MPA) and mean intersection over union (MIOU).