• Journal of KIISE:Software and Applications
• /
• v.33 no.11
• /
• pp.936-941
• /
• 2006

To apply object recognition techniques to real environment, illumination compensation method should be developed. As effective illumination compensation model, we focused our attention on Retinex model and illumination-Reflectance model, implemented them, and experimented on their performance. We implemented Retinex model with Single Scale Retinex, Multi-Scale Retinex, and Retinex Neural Network and Multi-Scale Retinex Neural Network, neural network model of Retinex model. Also, we implemented illumination-Reflectance model with reflectance image calculation by calculating an illumination image by low frequency filtering in frequency domain of Discrete Cosine Transform and Wavelet Transform, and Gaussian blurring. We compare their illumination compensation performance to facial images under nine illumination directions. We also compare their performance after post processing using Principal Component Analysis(PCA). As a result, illumination Reflectance model showed better performance and their overall performance was improved when illumination compensated images were post processed by PCA.

• Journal of IKEEE
• /
• v.26 no.3
• /
• pp.468-474
• /
• 2022

In this paper, we propose a method to implement a real-time fire alarm system using deep learning. The deep learning image dataset for fire alarms acquired 1,500 sheets through the Internet. If various images acquired in a daily environment are learned as they are, there is a disadvantage that the learning accuracy is not high. In this paper, we propose a fire image data expansion method to improve learning accuracy. The data augmentation method learned a total of 2,100 sheets by adding 600 pieces of learning data using brightness control, blurring, and flame photo synthesis. The expanded data using the flame image synthesis method had a great influence on the accuracy improvement. A real-time fire detection system is a system that detects fires by applying deep learning to image data and transmits notifications to users. An app was developed to detect fires by analyzing images in real time using a model custom-learned from the YOLO V4 TINY model suitable for the Edge AI system and to inform users of the results. Approximately 10% accuracy improvement can be obtained compared to conventional methods when using the proposed data.

• Journal of Information Processing Systems
• /
• v.19 no.3
• /
• pp.289-301
• /
• 2023

Due to the development and dissemination of modern technology, anyone can easily communicate using services such as social network service (SNS) through a personal computer (PC) or smartphone. The development of these technologies has caused many beneficial effects. At the same time, bad effects also occurred, one of which was the spam problem. Spam refers to unwanted or rejected information received by unspecified users. The continuous exposure of such information to service users creates inconvenience in the user's use of the service, and if filtering is not performed correctly, the quality of service deteriorates. Recently, spammers are creating more malicious spam by distorting the image of spam text so that optical character recognition (OCR)-based spam filters cannot easily detect it. Fortunately, the level of transformation of image spam circulated on social media is not serious yet. However, in the mail system, spammers (the person who sends spam) showed various modifications to the spam image for neutralizing OCR, and therefore, the same situation can happen with spam images on social media. Spammers have been shown to interfere with OCR reading through geometric transformations such as image distortion, noise addition, and blurring. Various techniques have been studied to filter image spam, but at the same time, methods of interfering with image spam identification using obfuscated images are also continuously developing. In this paper, we propose a deep learning-based spam image detection model to improve the existing OCR-based spam image detection performance and compensate for vulnerabilities. The proposed model extracts text features and image features from the image using four sub-models. First, the OCR-based text model extracts the text-related features, whether the image contains spam words, and the word embedding vector from the input image. Then, the convolution neural network-based image model extracts image obfuscation and image feature vectors from the input image. The extracted feature is determined whether it is a spam image by the final spam image classifier. As a result of evaluating the F1-score of the proposed model, the performance was about 14 points higher than the OCR-based spam image detection performance.

• Progress in Medical Physics
• /
• v.24 no.1
• /
• pp.76-83
• /
• 2013

Previous studies about effect of respiratory motion on diagnostic imaging and radiation therapy have been performed by monitoring external motions but these can not reflect internal organ motion well. The aim of this study was to develope the artificial pulmonary nodule able to perform non-invasive implantation to dogs in the thorax and to evaluate applicability of the model to respiratory motion studies on PET image acquisition and radiation delivery by phantom studies. Artificial pulmonary nodule was developed on the basis of 8 Fr disposable gastric feeding tube. Four anesthetized dogs underwent implantation of the models via trachea and implanted locations of the models were confirmed by fluoroscopic images. Artificial pulmonary nodule models for PET injected $^{18}F$-FDG and mounted on the respiratory motion phantom. PET images of those acquired under static, 10-rpm- and 15-rpm-longitudinal round motion status. Artificial pulmonary nodule models for radiation delivery inserted glass dosemeter and mounted on the respiratory motion phantom. Radiation delivery was performed at 1 Gy under static, 10-rpm- and 15-rpm-longitudinal round motion status. Fluoroscpic images showed that all models implanted in the proximal caudal bronchiole and location of models changed as respiratory cycle. Artificial pulmonary nodule model showed motion artifact as respiratory motion on PET images. SNR of respiratory gated images was 7.21. which was decreased when compared with that of reference images 10.15. However, counts of respiratory images on profiles showed similar pattern with those of reference images when compared with those of static images, and it is assured that reconstruction of images using by respiratory gating improved image quality. Delivery dose to glass dosemeter inserted in the models were same under static and 10-rpm-longitudinal motion status with 0.91 Gy, but dose delivered under 15-rpm-longitudinal motion status was decreased with 0.90 Gy. Mild decrease of delivered radiation dose confirmed by electrometer. The model implanted in the proximal caudal bronchiole with high feasibility and reflected pulmonary internal motion on fluoroscopic images. Motion artifact could show on PET images and respiratory motion resulted in mild blurring during radiation delivery. So, the artificial pulmonary nodule model will be useful tools for study about evaluation of motion on diagnostic imaging and radiation therapy using laboratory animals.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.45 no.4
• /
• pp.41-49
• /
• 2008

H.264/AVC supports variable block motion compensation, multiple reference images, 1/4-pixel motion vector accuracy, and in-loop deblocking filter, compared with the existing compression technologies. While these coding technologies are major functions of compression rate improvement, they lead to high complexity at the same time. For the H.264 video coding technology to be actually applied on low-end / low-bit rates terminals more extensively, it is essential to improve tile coding speed. Currently the deblocking filter that can improve the moving picture's subjective image quality to a certain degree is used on low-end terminals to a limited extent due to computational complexity. In this paper, a performance improvement method of the deblocking filter that efficiently reduces the blocking artifacts occurred during the compression of low-bit rates digital motion pictures is suggested. In the method proposed in this paper, the image's spatial correlational characteristics are extracted by using the variable block information of motion compensation; the filtering is divided into 4 modes according to the characteristics, and adaptive filtering is executed in the divided regions. The proposed deblocking method reduces the blocking artifacts, prevents excessive blurring effects, and improves the performance about $30{\sim}40%$ compared with the existing method.

• The Korean Journal of Nuclear Medicine Technology
• /
• v.21 no.2
• /
• pp.55-64
• /
• 2017

Purpose Recently, with the spread of SPECT/CT, various image correction methods can be applied quickly and accurately, which enabled us to expect quantitative accuracy as well as image quality improvement. Among them, the Collimator Detector Response(CDR) recovery is a correction method aiming at resolution recovery by compensating the blurring effect generated from the distance between the detector and the object. The purpose of this study is to find out quantitative change depending on the change in detection distance in SPECT/CT images with CDR recovery applied. Materials and Methods In order to find out the error of acquisition count depending on the change of detection distance, we set the detection distance according to the obit type as X, Y axis radius 30cm for circular, X, Y axis radius 21cm, 10cm for non-circular and non-circular auto(=auto body contouring, ABC_spacing limit 1cm) and applied reconstruction methods by dividing them into Astonish(3D-OSEM with CDR recovery) and OSEM(w/o CDR recovery) to find out the difference in activity recovery depending on the use of CDR recovery. At this time, attenuation correction, scatter correction, and decay correction were applied to all images. For the quantitative evaluation, calibration scan(cylindrical phantom, $^{99m}TcO_4$ 123.3 MBq, water 9293 ml) was obtained for the purpose of calculating the calibration factor(CF). For the phantom scan, a 50 cc syringe was filled with 31 ml of water and a phantom image was obtained by setting $^{99m}TcO_4$ 123.3 MBq. We set the VOI(volume of interest) in the entire volume of the syringe in the phantom image to measure total counts for each condition and obtained the error of the measured value against true value set by setting CF to check the quantitative accuracy according to the correction. Results The calculated CF was 154.28 (Bq/ml/cps/ml) and the measured values against true values in each conditional image were analyzed to be circular 87.5%, non-circular 90.1%, ABC 91.3% and circular 93.6%, non-circular 93.6%, ABC 93.9% in OSEM and Astonish, respectively. The closer the detection distance, the higher the accuracy of OSEM, and Astonish showed almost similar values regardless of distance. The error was the largest in the OSEM circular(-13.5%) and the smallest in the Astonish ABC(-6.1%). Conclusion SPECT/CT images showed that when the distance compensation is made through the application of CDR recovery, the detection distance shows almost the same quantitative accuracy as the proximity detection even under the distant condition, and accurate correction is possible without being affected by the change in detection distance.

• Progress in Medical Physics
• /
• v.27 no.1
• /
• pp.1-7
• /
• 2016

In this work, we considered a compressed-sensing (CS)-based image deblurring scheme with a total-variation (TV) regularization penalty for improving image characteristics in digital tomosynthesis (DTS). We implemented the proposed image deblurring algorithm and performed a systematic simulation to demonstrate its viability. We also performed an experiment by using a table-top setup which consists of an x-ray tube operated at $90kV_p$, 6 mAs and a CMOS-type flat-panel detector having a $198-{\mu}m$ pixel resolution. In the both simulation and experiment, 51 projection images were taken with a tomographic angle range of ${\theta}=60^{\circ}$ and an angle step of ${\Delta}{\theta}=1.2^{\circ}$ and then deblurred by using the proposed deblurring algorithm before performing the common filtered-backprojection (FBP)-based DTS reconstruction. According to our results, the image sharpness of the recovered x-ray images and the reconstructed DTS images were significantly improved and the cross-plane spatial resolution in DTS was also improved by a factor of about 1.4. Thus the proposed deblurring scheme appears to be effective for the blurring problems in both conventional radiography and DTS and is applicable to improve the present image characteristics.

• Journal of Korea Multimedia Society
• /
• v.13 no.3
• /
• pp.438-456
• /
• 2010

Iris recognition uses a unique iris pattern of user to identify person. In order to enhance the performance of iris recognition, it is reported that the diameter of iris region should be greater than 200 pixels in the captured iris image. So, the previous iris system used zoom lens camera, which can increase the size and cost of system. To overcome these problems, we propose a new method of enhancing the accuracy of iris recognition on low-resolution iris images which are captured without a zoom lens. This research is novel in the following two ways compared to previous works. First, this research is the first one to analyze the performance degradation of iris recognition according to the decrease of the image resolution by excluding other factors such as image blurring and the occlusion of eyelid and eyelash. Second, in order to restore a high-resolution iris image from single low-resolution one, we propose a new method based on multiple multi-layered perceptrons (MLPs) which are trained according to the edge direction of iris patterns. From that, the accuracy of iris recognition with the restored images was much enhanced. Experimental results showed that when the iris images down-sampled by 6% compared to the original image were restored into the high resolution ones by using the proposed method, the EER of iris recognition was reduced as much as 0.133% (1.485% - 1.352%) in comparison with that by using bi-linear interpolation

• Journal of Korean Academy of Oral and Maxillofacial Radiology
• /
• v.21 no.2
• /
• pp.317-326
• /
• 1991

The purpose of this study was to establish the principle and the clinical application of dynamic tomogram of a human head by using the dental machine. For this study, a block of wax with details lying at three parallel planes and a human dry skull were used. This experiment was reexamined the dynamic tomogram with specialized radiographic device and view box, and the radiograms taken by the change of exposure time according to the numbers of film used in x-ray taking and taken according to the change of kVp and the types of film were analyzed density with the densitometer. From this study, the obtained results were as follows: 1. When the underexposed radiograms taken by angulation of clockwise and counter-clockwise direction of the film and skull. were superimposed and moved laterally, it was possible to focus on right and left jaws and teeth. 2. The superimposition of the two underexposed radiograms according to each condition of x-ray taking showed some differencies in density visually, and the measurement of density with the densitometer was 1.23 to 1.57 in 75kVp and 1.34 to 1.70 in 90kVp. 3. The superimposition of the two underexposed radiograms according to the kinds of x-ray film showed almost equal density visually, and the measurement of density with the densiometer was 1.34 to 1.37. 4. When seven radiograms taken by each condition of x-ray taking were superimposed on the view box, a intense rear light of view box didn't transilluminate film density regardless of the conditions of x-ray taking. Even though seven radiograms taken according to types of film were superimposed on the view box, a more intense rear light of view box was required to transilluminate total density of films. 6. Long film-object distance resulted in the enlargement and blurring of radiographic images.

• 한국정보디스플레이학회:학술대회논문집
• /
• 2004.08a
• /
• pp.843-846
• /
• 2004

For the fast growing Liquid Crystal Display (LCD) TV market, it is essential to make the LCD panels to show moving images without any visual difficulties such as blurring or tailing. Owing to reduction of the cell gap and the improved Liquid Crystal (LC) mixtures with low viscosity, it is possible that our S-IPS TFT-LCDs feature a response time (R/T) as fast as 1-frame time (16ms) for a white-black operation and less than a 16rns in all gray levels without Over Driving Circuit (ODC) technology. Currently, mass production of the large size IPS panels with high speed has been successfully achieved. In order to achieve faster response time, new LC mixtures have been developed, optimizing the physical properties of rotational viscosity (${\gamma}$1) and elastic constants (Kii). Also, the LC mixtures with high elastic constant allow us to increase the cell gap. In this paper, realization of fast switching time in IPS mode with optimized '${\gamma}$1/Kii' parameter in the LC mixtures forms the core of this paper.