• Journal of Broadcast Engineering
• /
• v.25 no.2
• /
• pp.200-207
• /
• 2020

Recently, deep neural networks such as CNN are showing excellent performance in various fields such as image classification, object recognition, visual quality enhancement, etc. However, as the model size and computational complexity of deep learning models for most applications increases, it is hard to apply neural networks to IoT and mobile environments. Therefore, neural network compression algorithms for reducing the model size while keeping the performance have been being studied. In this paper, we apply few compression methods to CNN models and evaluate their performances in the embedded environment. For evaluate the performance, the classification performance and inference time of the original CNN models and the compressed CNN models on the image inputted by the camera are evaluated in the embedded board equipped with QCS605, which is a customized AI chip. In this paper, a few CNN models of MobileNetV2, ResNet50, and VGG-16 are compressed by applying the methods of pruning and matrix decomposition. The experimental results show that the compressed models give not only the model size reduction of 1.3~11.2 times at a classification performance loss of less than 2% compared to the original model, but also the inference time reduction of 1.2~2.21 times, and the memory reduction of 1.2~3.8 times in the embedded board.

• Journal of the Society of Naval Architects of Korea
• /
• v.54 no.5
• /
• pp.406-414
• /
• 2017

An experimental study in a recirculating water channel was carried out to investigate the effect of large coherent structures to the skin friction on a flat plate. Particle Image Velocimetry (PIV) technique was used to quantify characteristic features of coherent structures growing to the boundary layer. In the PIV measurement, it is difficult to calculate the friction velocity near the wall region due to laser deflection and uncertainty so that Clauser fitting method at the logarithmic region was adopted to compute the friction velocity and compared with the one directly measured by the dynamometer. With changing the free-stream velocity from 0.5 m/s to 1.0 m/s, the activity of coherent structures in the logarithmic region was increased over three times in terms of Reynolds stress. The flow field was separated by Variable Interval Time Averaging (VITA) technique into the weak and the strong structure case depending on the existence large coherent structures in order to validate its effectiveness. The stream-wise velocity fluctuation was scanned through at the boundary thickness whether it had a large deviation from background flow. With coherent structures connected from near-wall to the boundary layer, mean wall shear stress was higher than that of weak structure case. Proper Orthogonal Decomposition (POD) analysis was also applied to compare the energy budget between them at each free-stream velocity.

• Journal of radiological science and technology
• /
• v.39 no.2
• /
• pp.227-236
• /
• 2016

Dynamic positron emission tomography(dPET) is widely used medical imaging modality that can provide both physiological and functional neuro-image for diagnosing various brain disease. However, dPET images have low spatial-resolution and high noise level during spatio-temporal analysis (three-dimensional spatial information + one-dimensional time information), there by limiting clinical utilization. In order to overcome these issues for the spatio-temporal analysis, a novel computational technique was introduced in this paper. The computational technique based on singular value decomposition classifies multiple independent components. Signal components can be distinguished from the classified independent components. The results show that signal to noise ratio was improved up to 30% compared with the original images. We believe that the proposed computational technique in dPET can be useful tool for various clinical / research applications.

• Korean Journal of Remote Sensing
• /
• v.17 no.1
• /
• pp.33-44
• /
• 2001

The paper presents several satellite models and satellite image decomposition methods for inaccessible area where ground control points can hardly acquired in conventional ways. First, 10 different satellite sensor models, which were extended from collinearity condition equations, were developed and then behavior of each sensor model was investigated. Secondly, satellite images were decomposed and also pseudo images were generated. The satellite sensor model extended from collinearity equations was represented by the six exterior orientation parameters in $1^{st}$, $2^{nd}$ and $3^{rd}$ order function of satellite image row. Among them, the rotational angle parameters such as $\omega$(omega) and $\Phi$(phi) correlated highly with positional parameters could be assigned to constant values. For inaccessible area, satellite images were decomposed, which means that two consecutive images were combined as one image, The combined image consists of one satellite image with ground control points and the other without ground control points. In addition, a pseudo image which is an imaginary image, was prepared from one satellite image with ground control points and the other without ground control points. In other words, the pseudo image is an arbitrary image bridging two consecutive images. For the experiments, SPOT satellite images exposed to the similar area in different pass were used. Conclusively, it was found that 10 different satellite sensor models and 5 different decomposed methods delivered different levels of accuracy. Among them, the satellite camera model with 1st order function of image row for positional orientation parameters and rotational angle parameter of kappa, and constant rotational angle parameter omega and phi provided the best 60m maximum error at check point with pseudo images arrangement.

• Applied Chemistry for Engineering
• /
• v.34 no.1
• /
• pp.80-85
• /
• 2023

New yellow quinoline-dione dye derivatives were designed and synthesized for use in image sensor color filters. The synthesized compounds have a basic chemical structure composed of quinoline and dione groups. New materials were evaluated on the basis of their optical and thermal properties under conditions mimicking those of a commercial device fabrication process. A comparison of their related performances revealed that, between the two prepared compounds, 2-(3-hydroxyquinolin-2(1H)-ylidene)-1H-indene-1,3(2H)-dione (HQIDO) exhibited the superior performance as an image sensor color filter material, including a solubility greater than 0.5 wt% in propylene glycol monomethyl ether acetate solvent and a high decomposition temperature of 298 ℃, respectively. The results suggest that HQIDO can be used as a yellow dye additive in an image sensor colorant.

• Applied Chemistry for Engineering
• /
• v.34 no.6
• /
• pp.590-595
• /
• 2023

Novel aromatic imine derivatives with yellow were designed and synthesized for their potential application in color filters for image sensors. The synthesized compounds possessed chemical structures using aromatic imine groups. This innovative material was evaluated thoroughly, considering its optical and thermal properties under conditions similar to commercial device manufacturing processes. Following a rigorous performance evaluation, it was found that (E)-3-methyl-4-((3-methyl-5-oxo-1-phenyl-1H-pyrazol-4(5H)-ylidene)methyl)-1-phenyl-1H-pyrazol-5(4H)-one, abbreviated as MOPMPO, exhibited an impressive solubility of 0.5 wt% in propylene glycol monomethyl ether acetate, predominantly utilized as the solvent in the industry. Furthermore, MOPMPO showed exceptional performance as a color filter material for image sensors, having a high decomposition temperature of 290 ℃. These data unequivocally establish MOPMPO as a viable yellow dye additive for coloring materials in image sensor applications.

• Proceedings of the KSRS Conference
• /
• 2001.03a
• /
• pp.140-148
• /
• 2001

The satellite sensor model is typically established using ground control points acquired by ground survey Of existing topographic maps. In some cases where the targeted area can't be accessed and the topographic maps are not available, it is difficult to obtain ground control points so that geospatial information could not be obtained from satellite image. The paper presents several satellite sensor models and satellite image decomposition methods for non-accessible area where ground control points can hardly acquired in conventional ways. First, 10 different satellite sensor models, which were extended from collinearity condition equations, were developed and then the behavior of each sensor model was investigated. Secondly, satellite images were decomposed and also pseudo images were generated. The satellite sensor model extended from collinearity equations was represented by the six exterior orientation parameters in 1$^{st}$, 2$^{nd}$ and 3$^{rd}$ order function of satellite image row. Among them, the rotational angle parameters such as $\omega$(omega) and $\phi$(phi) correlated highly with positional parameters could be assigned to constant values. For non-accessible area, satellite images were decomposed, which means that two consecutive images were combined as one image. The combined image consists of one satellite image with ground control points and the other without ground control points. In addition, a pseudo image which is an imaginary image, was prepared from one satellite image with ground control points and the other without ground control points. In other words, the pseudo image is an arbitrary image bridging two consecutive images. For the experiments, SPOT satellite images exposed to the similar area in different pass were used. Conclusively, it was found that 10 different satellite sensor models and 5 different decomposed methods delivered different levels of accuracy. Among them, the satellite camera model with 1$^{st}$ order function of image row for positional orientation parameters and rotational angle parameter of kappa, and constant rotational angle parameter omega and phi provided the best 60m maximum error at check point with pseudo images arrangement.

• Journal of Physiology & Pathology in Korean Medicine
• /
• v.23 no.1
• /
• pp.97-103
• /
• 2009

The status and the property of a tongue are the important indicators to diagnose one's health like physiological and clinicopathological changes of inner organs. However, the tongue diagnosis is affected by examination circumstances like a light source, patient's posture, and doctor's condition. To develop an automatic tongue diagnosis system for an objective and standardized diagnosis, classifying tongue coating is inevitable but difficult since the features like color and texture of the tongue coatings and substance have little difference, especially in the neighborhood on the tongue surface. The proposed method has two procedures; the first is to acquire the color table to classify tongue coatings and substance by automatically separating coating regions marked by oriental medical doctors, decomposing the color components of the region into hue, saturation and brightness and obtaining the 2nd order discriminant with statistical data of hue and saturation corresponding to each kind of tongue coatings, and the other is to apply the tongue region in an input image to the color table, resulting in separating the regions of tongue coatings and classifying them automatically. As a result, kinds of tongue coatings and substance were segmented from a face image corresponding to regions marked by oriental medical doctors and the color table for classification took hue and saturation values as inputs and produced the classification of the values into white coating, yellow coating and substance in a digital tongue diagnosis system. The coating regions classified by the proposed method were almost the same to the marked regions. The exactness of classification was 83%, which is the degree of correspondence between what Oriental medical doctors diagnosed and what the proposed method classified. Since the classified regions provide effective information, the proposed method can be used to make an objective and standardized diagnosis and applied to an ubiquitous healthcare system. Therefore, the method will be able to be widely used in Oriental medicine.

• Journal of Broadcast Engineering
• /
• v.2 no.2
• /
• pp.146-155
• /
• 1997

The new video coding standard Iv1PEG-4 is enabling content-based functionalities. It requires a prior decomposition of sequences into video object planes (VOP's) so that each VOP represents moving objets. This paper addresses an image segmentation method for separating moving objects from still background (non-moving area) in video sequences using a statistical hypothesis test. In the proposed method. three consecutive image frames are exploited and a hypothesis testing is performed by comparing two means from two consecutive difference images. which results in a T-test. This hypothesis test yields a change detection mask that indicates moving areas (foreground) and non-moving areas (background), Moreover. an effective method for extracting

• Journal of Korea Multimedia Society
• /
• v.12 no.9
• /
• pp.1233-1241
• /
• 2009

Over the 50 percents of women who are older than 45 years have osteoporosis. Because people hardly recognize this disease by themselves, the researches that measure bone mineral density have been doing widely to detect osteoporosis in the early stage. The most widely used methods for bone mineral density measurement are based on the X-ray imaging. Among them, DEXA(Dual-energy X-ray Absorptiometry) imaging is one of the important methods in bone mineral density measurement. DEXA images are useful methods to increase diagnosis efficiency by reducing anatomic noise as two images obtained from two different energy levels. However, it has some problems to a calibration parameter determined by the heuristic method for bone extraction. In this paper, we propose the method to extract bone in DEXA image using calibration parameter based on anatomic attenuation coefficient. The experimental results reveal that the proposed method is effective.