• Journal of Korea Multimedia Society
• /
• v.21 no.12
• /
• pp.1387-1395
• /
• 2018

Core algorithm of deep learning Convolutional Neural Network(CNN) shows better performance than other machine learning algorithms. However, if there is not sufficient data, CNN can not achieve satisfactory performance even if the classifier is excellent. In this situation, it has been proven that the use of transfer learning can have a great effect. In this paper, we apply two transition learning methods(freezing, retraining) to three CNN models(ResNet-50, Inception-V3, DenseNet-121) and compare and analyze how the classification performance of CNN changes according to the methods. As a result of statistical significance test using various evaluation indicators, ResNet-50, Inception-V3, and DenseNet-121 differed by 1.18 times, 1.09 times, and 1.17 times, respectively. Based on this, we concluded that the retraining method may be more effective than the freezing method in case of transition learning in image classification problem.

• Journal of the Korea Computer Graphics Society
• /
• v.24 no.4
• /
• pp.1-9
• /
• 2018

We propose an emotional training framework that can determine the initial symptom of schizophrenia by using emotional analysis method through facial expression change. We use Emotion API in Microsoft to obtain facial expressions and emotion values at the present time. We analyzed these values and recognized subtle facial expressions that change with time. The emotion states were classified according to the peak analysis-based variance method in order to measure the emotions appearing in facial expressions according to time. The proposed method analyzes the lack of emotional recognition and expressive ability by using characteristics that are different from the emotional state changes classified according to the six basic emotions proposed by Ekman. As a result, the analyzed values are integrated into the image color transfer framework so that users can easily recognize and train their own emotional changes.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.30 no.12C
• /
• pp.1214-1221
• /
• 2005

In this paper, we propose a digital watermarking method for still images, in which the human visual system (HVS) is used in the discrete cosine transform (DCT) domain. The modulation transfer function (MTF) of the HVS model is employed to increase the invisibility of the inserted watermark in images. The proposed watermarking method is shown to be robust to several common image processing techniques, including lowpass filtering and cropping. Also, using the energy relationship of the DCT, we derive the equation that directly computes the watermark weighting factor in the DCT domain for the specified peak signal to noise ratio (PSNR) of the still image and the length of watermark to be inserted. The difference between desired PSNR and PSNR in spatial domain is within 0.07dB for the 7 test images.

• Journal of Korea Multimedia Society
• /
• v.24 no.8
• /
• pp.1000-1011
• /
• 2021

In recent years, using Deep Learning methods to apply for medical and biomedical image analysis has seen many advancements. In clinical, using Deep Learning-based approaches for cancer image analysis is one of the key applications for cancer detection and treatment. However, the scarcity and shortage of labeling images make the task of cancer detection and analysis difficult to reach high accuracy. In 2015, the Unet model was introduced and gained much attention from researchers in the field. The success of Unet model is the ability to produce high accuracy with very few input images. Since the development of Unet, there are many variants and modifications of Unet related architecture. This paper proposes a new approach of using Unet++ with pretrained EfficientNet as backbone architecture for breast tumor cell nuclei segmentation and uses the multi-organ transfer learning approach to segment nuclei of breast tumor cells. We attempt to experiment and evaluate the performance of the network on the MonuSeg training dataset and Triple Negative Breast Cancer (TNBC) testing dataset, both are Hematoxylin and Eosin (H & E)-stained images. The results have shown that EfficientUnet++ architecture and the multi-organ transfer learning approach had outperformed other techniques and produced notable accuracy for breast tumor cell nuclei segmentation.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.19 no.9
• /
• pp.2386-2398
• /
• 1995

The paper describes the results of an experimental investigation of the heat transfer rate on the endwall surface within the plane turbine cascade passage and includes the effect of the heat transfer for the two different boundary layer thicknesses and Reynolds numbers. The limiting streamlines on the endwall surface have been visualized by the oil film method in order to compare with the endwall heat transfer. The hue-capturing method using the termochromatic liquid crystals with great spatial resolution has been used to provide the local distribution of the endwall heat transfer coefficients. Because the detailed contours of the local heat transfer coefficients over the entire endwall can be obtained from the hue-capturing method, it has been possible to obtain information on the endwall heat transfer within the plane turbine cascade passage from these heat transfer contours.

• Journal of Astronomy and Space Sciences
• /
• v.24 no.4
• /
• pp.379-388
• /
• 2007

Modulation Transfer Function (MTF) of satellite image is an important performance index in satellite image applications. Therefore MTF performance is assessed using satellite image for the ground target during LEOP phase after launch. But the MTF performance assessment using the ground target can be affected by imaging conditions such as cloud and weather. In this paper system requirements and satellite operation for assessing MTF performance of satellite image using stellar sources are proposed. Satellite capability in collecting stellar sources using the satellite which is designed for earth observation and satellite image usefulness for assessing MTF performances were analyzed. The proposed approach will be useful to assess MTF performance of earth observation satellite in lower earth orbit.

• Proceedings of the KOSOMBE Conference
• /
• v.1990 no.11
• /
• pp.42-47
• /
• 1990

In this paper, a new adaptive image processing method based on human visual system has been presented. The basic idea behind the proposed method is to improve the efficiency of the information transfer channel regionally by manipulating the displayed image in order to compensate the regional inefficiency of the information transfer channel. The proposed method consists of two parts; the first part reallocates pixel values corresponding to high X-ray attenuation to that of more intense X-ray exposure by multiplying the pixel values with the local adaptive multiplcation factor, and the second part adjusts the pixel values of dark area of displayed image such as overexposed lung area to be more bright. The processed image with the proposed method shows significantly increased visibility of mediastinal and subdiaphramatic area, and also the lung area of over exposed case without any artifact.

• Fire Science and Engineering
• /
• v.26 no.2
• /
• pp.69-74
• /
• 2012

Thermal imaging technology based on IR sensor with high transmittance through the fire smoke is considered as one of the advanced technology for the fire service. The present study has been performed to investigate the image saturation characteristics with thermal condition of target and background and evaluate the performance of image quality based on the contrast transfer function (CTF). For the present testing conditions, TIC using BST sensor did not show the image saturation and the image quality based on the CTF was proportional to the temperature difference between target and background. This study can be utilized as preliminary study to improve reliability and technical development of TIC.

• Aerospace Engineering and Technology
• /
• v.1 no.2
• /
• pp.66-74
• /
• 2002

The characteristics of the satellite image quality parameters driven by satellite imaging instrument are investigated. Since the satellite image is directly produced by the satellite imaging instrument, the satellite image quality depends on the imager performance highly. This is why the imager performance parameters are considered as an important part of the satellite image quality parameters. The imager performance parameters consist of spectral band parameters, ground sample distance(GSD) parameters, swath width parameters, imager Modulation Transfer Function (MTF), imager Signal to Noise Ratio(SNR), radiometric response characteristics parameters, pixel registration, and imager calibration.

• Proceedings of the Korean Institute of Electrical and Electronic Material Engineers Conference
• /
• 2003.11a
• /
• pp.384-387
• /
• 2003

In this paper, the evaluation of image quality was performed for digital radiography which is developing in using amorphous selenium as a photoconductor material for the purpose of offering basic research data and measurement technique about Medical Imaging Quality. So Modulation Transfer Function as a main factor of imaging quality evaluation was investigated by slit method. For measurement of MTF, Nuclear associates. 07-624 Slit camera image was obtained to study the variation of MTF corresponding to changing spatial frequency. And Presampling MTF was estimated by slit camera image with $10\;{\mu}m$ width at Digital Radiography. In this study, the obtained data demonstrates that the clinical value of a direct conversion type digital radiation detector using the amorphous selenium, which is being developed by domestic technology.