• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.21 no.7
• /
• pp.1677-1684
• /
• 1996

The theory of optimal stack filtering has been used in difference of estimates(DoE) approach to the detection of intensity edges in noisy image. In this approach, stack filters are applied to a noisy image to obtain local estimates of the dilated and eroded versions of the noise-free image. Thresholding the difference between these two estimates produces the estimated edge map. In this paper, the DoE approach is modified by imposing a symmetry condition of the data used to train the two stack filers. Under this condition, the stack filters obtained are duals of each other. Only one filter must therefore be trained;the other is simply its dual. They also produce statistially unbiased estimates. This new technique is called the symmetric Difference of Estimates (SDoE) approach.

• Annual Conference of KIPS
• /
• 2023.11a
• /
• pp.557-560
• /
• 2023

In this paper, we introduce a pre-training method leveraging the capabilities of the Vision Transformer (ViT) for disease diagnosis in conventional Fundus images. Recognizing the need for effective representation learning in medical images, our method combines the Vision Transformer with a Masked Autoencoder to generate meaningful and pertinent image augmentations. During pre-training, the Masked Autoencoder produces an altered version of the original image, which serves as a positive pair. The Vision Transformer then employs contrastive learning techniques with this image pair to refine its weight parameters. Our experiments demonstrate that this dual-model approach harnesses the strengths of both the ViT and the Masked Autoencoder, resulting in robust and clinically relevant feature embeddings. Preliminary results suggest significant improvements in diagnostic accuracy, underscoring the potential of our methodology in enhancing automated disease diagnosis in fundus imaging.

• ETRI Journal
• /
• v.46 no.3
• /
• pp.513-525
• /
• 2024

Deep neural networks trained on labeled medical data face major challenges owing to the economic costs of data acquisition through expensive medical imaging devices, expert labor for data annotation, and large datasets to achieve optimal model performance. The heterogeneity of diseases, such as Alzheimer's disease, further complicates deep learning because the test cases may substantially differ from the training data, possibly increasing the rate of false positives. We propose a reconstruction-based self-supervised anomaly detection model to overcome these challenges. It has a dual-subnetwork encoder that enhances feature encoding augmented by skip connections to the decoder for improving the gradient flow. The novel encoder captures local and global features to improve image reconstruction. In addition, we introduce an entropy-based image conversion method. Extensive evaluations show that the proposed model outperforms benchmark models in anomaly detection and classification using an encoder. The supervised and unsupervised models show improved performances when trained with data preprocessed using the proposed image conversion method.

• Journal of radiological science and technology
• /
• v.45 no.2
• /
• pp.135-143
• /
• 2022

In SPECT image, scatter count is the cause of quantitative count error and image quality degradation. This study is to evaluate the accuracy of CT based SC(CTSC) and energy window based SC(EWSC) as the comparison with existing Non SC. SPECT/CT images were obtained after filling air in order to acquire a reference image without the influence of scatter count inside the Triple line insert phantom setting hot rod(^99mTc 74.0 MBq) in the middle and each SPECT/CT image was obtained each separately after filling water instead of air in order to derive the influence of scatter count under the same conditions. For EWSC, 9 sub-energy windows were set additionally in addition to main energy window(140 keV, 20%) and then, images were acquired at the same time and five types of EWSC including DPW(dual photo-peak window)10%, DEW(dual energy window)20%, TEW(triple energy window)10%, TEW5.0%, TEW2.5% were used. Under the condition without fluctuations in primary count, total count was measured by drawing volume of interest (VOI) in the images of the two conditions and then, the ratio of scatter count of total counts was calculated as percent scatter fraction(%SF) and the count error with image filled with water was evaluated with percent normalized mean-square error(%NMSE) based on the image filled with air. Based on the image filled with air, %SF of images filled with water to which each SC method was applied is non scatter correction(NSC) 37.44, DPW 27.41, DEW 21.84, TEW10% 19.60, TEW5% 17.02, TEW2.5% 14.68, CTSC 5.57 and the scatter counts were removed the most in CTSC and %NMSE is NSC 35.80, DPW 14.28, DEW 7.81, TEW10% 5.94, TEW5% 4.21, TEW2.5% 2.96, CTSC 0.35 and the error in CTSC was found to be the lowest. In SPECT/CT images, the application of each scatter correction method used in the experiment could improve the quantitative count error caused by the influence of scatter count. In particular, CTSC showed the lowest %NMSE(=0.35) compared to existing EWSC methods, enabling relatively accurate scatter correction.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.11 no.9
• /
• pp.3422-3427
• /
• 2010

In this paper, we presents a sensor calibration technique using stereo X-ray images to provide efficient inspection of fast moving cargo objects. Stereo X-ray scanned images are acquired from a specially designed equipment which consists of a X-ray source, dual-linear array detector, and a conveyor system. Dual detector is installed so that rectified stereo X-ray images of objects are acquired. Using the stereo X-ray images, we carry out a sensor calibration to find the correspondences between the images and reconstruct 3-D shapes of real objects. Using the Image acquired from the stereo detectors with varying distances, we calculated the GCP(ground control point)of the image. And we figure out the error by comparing calculated GCP and GCP of the real object. The experimental results show the proposed technique can enhance the accuracy of stereo matching and give more efficient visualization for cargo inspection image.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.43 no.5 s.311
• /
• pp.52-61
• /
• 2006

Orthogonal wavelet tansform which is generally used in image and signal processing applications has limited performance because of lack of shift invariance and low directional selectivity. To overcome these demerits complex wavelet transform has been proposed. In this paper, we present an efficient image denoising method using dual-tree complex wavelet transform and Bernoulli-Gauss prior model. In estimating hyper-parameters for Bernoulli-Gaussian model, we present two simple and non-iterative methods. We use hypothesis-testing technique in order to estimate the mixing parameter, Bernoulli random variable. Based on the estimated mixing parameter, variance for clean signal is obtained by using maximum generalized marginal likelihood (MGML) estimator. We simulate our denoising method using dual-tree complex wavelet and compare our algorithm to well blown denoising schemes. Experimental results show that the proposed method can generate good denoising results for high frequency image with low computational cost.

• Journal of Sensor Science and Technology
• /
• v.24 no.2
• /
• pp.131-136
• /
• 2015

This paper presents a wide dynamic range complementary metal oxide semiconductor (CMOS) image sensor with column capacitor and feedback structure. The designed circuit has been fabricated by using $0.18{\mu}m$ 1-poly 6-metal standard CMOS technology. This sensor has dual mode operation using combination of active pixel sensor (APS) and passive pixel sensor (PPS) structure. The proposed pixel operates in the APS mode for high-sensitivity in normal light intensity, while it operates in the PPS mode for low-sensitivity in high light intensity. The proposed PPS structure is consisted of a conventional PPS with column capacitor and feedback structure. The capacitance of column capacitor is changed by controlling the reference voltage using feedback structure. By using the proposed structure, it is possible to store more electric charge, which results in a wider dynamic range. The simulation and measurement results demonstrate wide dynamic range feature of the proposed PPS.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2014.05a
• /
• pp.695-698
• /
• 2014

As most the scanning systems developed until now provide radiation scan plane images of the inspected objects, there has been a limitation in judging exactly the shape of the objects inside a logistics container exactly with only 2-D radiation image information. Two 2-dimensional radiation images which have different disparity values are acquired from a newly designed stereo image acquisition system which has one additional line sensor to the conventional system. Using a matching algorithm the 3D reconstruction process which find the correspondence between the images is progressed. In this paper, we proposed a new volume based 3D reconstruction algorithm and experimental results show the proposed new volume based reconstruction technique can provide more efficient visualization for cargo inspection. The proposed technique can be used for the development of the high speed and more efficient non-destructive auto inspection system.

• Journal of Institute of Control, Robotics and Systems
• /
• v.19 no.3
• /
• pp.240-247
• /
• 2013

This paper proposes an optimal three-dimensional coordinate implementation of the vision sensor using two CCD cameras. The PBVS (Position based visual servoing) is implemented using the positional information obtained from images. Stereo vision by PBVS method that has enhanced every frame using calibration parameters is effective in the distance calculation. The IBVS (Image based visual servoing) is also implemented using the difference between reference and obtained images. Stereo vision by IBVS method calculates the distance using rotation angle of motors that correspond eyes and neck without enhanced images. The PBVS method is compared with the IBVS method in terms of advantages, disadvantages, computing time, and performances. Finally, the IBVS method is applied for the dual arm manipulator on the mobile inverted pendulum. The autonomous mobile inverted pendulum is successfully demonstrated using the center of the manipulator's mass.

• Korean Journal of Optics and Photonics
• /
• v.8 no.4
• /
• pp.345-352
• /
• 1997

Recently, in designing optical mask such as 4GDRAM, the scattering effect of electromagnetic wave must be considered. For this reason we claculated directly the mask function using the finite difference time domain(FDTD) method. The modification of image theory with this new mask function could explain clearly the scattering effect at the etched side wall of the submicron optical mask. The characteristics of the various type of alternating PSM were investigated. According to the simulation, the dual wet etch process was the most useful fabrication technique to overcoe the light scattering off at the shifted opening.