• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2002년도 하계종합학술대회 논문집(4)
• /
• pp.145-148
• /
• 2002

A study on image searching and management techniques is actively developed by user requirements for multimedia information that are existing as images, audios, texts data from various information processing devices. We had been studied an automatical shape region segmentation method using color. distribution and edge characteristics of moving images for. contents-base description. The Proposed method uses a color information quantized on human visual system and extracts overlapped regions to be matched by using edge characteristics of the image frame. The performance of the proposed method is represented by similarity for comparison to a segmented image and original image.

• 제어로봇시스템학회논문지
• /
• 제6권7호
• /
• pp.544-550
• /
• 2000

Among the several types of 3D display methods the autostereoscopic method has an advantage that we can enjoy a 3D image without any additional device but the method has a disadvantage of a narrow viewing zone so that the moving viewer coannot see the 3D image continuously. This disadvantage can be overcome with the detectioni of viewer's positional movement by head tracking. In this paper we suggest a method of detecting the position of the moving viewer's two eyes by using images obtained through a color CCD camera, The suggested method consists of the preprocessing process and the eye-detection process. Through the experiment of applying the suggested method we were able to find the accurate two-eyes position for 78 images among 80 sample input images of 8 different men with the processing speed of 0.39 second/frame using a personal computer.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2000년도 ITC-CSCC -2
• /
• pp.889-893
• /
• 2000

In this paper, a method to detect lanes and obstacles from the images captured by a CCD camera fitted in an automobile is proposed, and a new terminology “Moving Window” is defined. Processing the input dynamic images in real time can cause quite a few constraints in terms of hardware. In order to overcome these problems and detect lanes and obstacles in real time using the images, the optimal size of “Moving Window” is determined, based upon road conditions and automobile states. The real time detection is made possible through the technique. For each image frame, the moving window is moved in a predicted direction, the accuracy of which is improved by the Kalman filter estimation. The feasibility of the proposed algorithm is demonstrated through the simulated experiments of freeway driving.

• 대한원격탐사학회지
• /
• 제24권3호
• /
• pp.251-256
• /
• 2008

Morphing method, one of classic image processing algorithms, has been used in various application fields. The motivation of this work is to investigate its applicability in consideration to geo-spatial data including airborne or space-borne images. For this purpose, the Beier and Neely morphing algorithm is tentatively implemented in the form of a prototype with user interface. As the results, this feature-based morphing with paired image sets can be used for general users: image simulation using two or more images and construction of color-blending image between source image and destination image in different types. Some simple application cases were demonstrated. This scheme is the simple and useful approach for those who want to utilize both geo-spatial data sets and airborne/space-borne image sets.

• ETRI Journal
• /
• 제26권3호
• /
• pp.218-228
• /
• 2004

Using stereo images with ephemeris data from the Korea Multi-Purpose Satellite-1 electro-optical camera (KOMPSAT-1 EOC), we performed geometric modeling for three-dimensional (3-D) positioning and evaluated its accuracy. In the geometric modeling procedures, we used ephemeris data included in the image header file to calculate the orbital parameters, sensor attitudes, and satellite position. An inconsistency between the time information of the ephemeris data and that of the center of the image frame was found, which caused a significant offset in satellite position. This time inconsistency was successfully adjusted. We modeled the actual satellite positions of the left and right images using only two ground control points and then achieved 3-D positioning using the KOMPSAT-1 EOC stereo images. The results show that the positioning accuracy was about 12-17 m root mean square error (RMSE) when 6.6 m resolution EOC stereo images were used along with the ephemeris data and only two ground control points (GCPs). If more accurate ephemeris data are provided in the near future, then a more accurate 3-D positioning will also be realized using only the EOC stereo images with ephemeris data and without the need for any GCPs.

• 한국의류학회지
• /
• 제24권1호
• /
• pp.128-137
• /
• 2000

A range of suede-like samples were collected including woven fabric type, nonwoven fabric type, and natural suede(sheep). The surface textures of these suede-like fabrics are rather diverse and different from the plain filament type fabrics since there are a lot of fine surface free fiber ends. Physical and mechanical measurements were carried using the KES equipments. Based on the Kawabata-Niwa translational equation, primary hand values and THV were calculated. Uniaxial tensile tests were performed. Using glossmeter, the reflectance pattern was analyzed at different incidence and receiving angles. In order to capture the surface images of the specimens, a CCD camera and frame grabber connected to a PC were employed. The reflectance uniformity of the images was measured with line-profile analysis and standard deviation values of the profile of the images were calculated. After the multiscale wavelet transfermation, correlation among the transformed image was analyzed at each scale. The reflectance uniformity of the natural suede was better than that of nonwoven type suede, while that of woven type suede was the last among the selected three samples(natural suede, nonwoven type, woven type). The correlation analysis among images has shown the possibility of using the wavelet transformation of the images as one of the measures to detect similarities among the textured specimens.

• 한국컴퓨터정보학회논문지
• /
• 제8권2호
• /
• pp.37-43
• /
• 2003

본 연구는 연속된 컬러 영상으로부터 전방의 차량과 차선을 검출하는 과정에서 연속 영상 분석을 통하여 다중 차량을 검출하는 방법을 제안한다. 하나의 프레임에서 차량 후보 영역의 검출은 그림자 특징과 에지 성분을 이용한다. 그리고, 다중 차량 영역을 검출하는 방법은 연속된 영상에 존재하는 차량 후보 영역들의 차량 추정값과(EOV)과 누적 유사도 함수(ASF)를 분석하여 차량일 가능성을 검사한다. 대부분의 연구 방법이 전방의 한 차량을 검출하는데 비해 본 연구에서는 여러 차량을 검출하는 방법을 제시하였으며, 교통량이 많고, 차선 변경이 자주 있는 경우에도 차량의 검출이 가능하도록 한다. 제안된 방식의 효과를 검증하기 위해 노트북 PC와 PC용 CCD 카메라로 도로에서의 영상을 촬영하고 차량 검출 알고리즘을 적용한 처리 시간, 정확도 및 차량검지 결과를 보인다.

• 한국측량학회지
• /
• 제20권3호
• /
• pp.331-341
• /
• 2002

평면 이미지로부터 3차원 정보를 구하기 위해서는 스테레오 이미지가 필요하고, 움직이는 물체의 거동을 3차원적으로 측정하기 위해서는 스테레오 이미지의 정확한 동기가 필수적이다. 또한, 연속적인 거동측정을 위해서는 동시 촬영한 연속 스테레오 이미지가 필요하고, 위치추적의 자동화를 위해서는 영상 매칭이 그 해결책이다. 본 연구에서는 한 대의 비디오 카메라로 움직이는 물체의 연속 스테레오 이미지를 촬영하여, 영상 매칭에 의해 동일 대상물 및 동일 측점의 중심위치를 결정하고, DLT에 의해 3차원 위치를 해석하는 방법을 제시하였다.

• Applied Microscopy
• /
• 제50권
• /
• pp.23.1-23.9
• /
• 2020

We propose an effective deep learning model to denoise scanning transmission electron microscopy (STEM) image series, named Noise2Atom, to map images from a source domain 𝓢 to a target domain 𝓒, where 𝓢 is for our noisy experimental dataset, and 𝓒 is for the desired clear atomic images. Noise2Atom uses two external networks to apply additional constraints from the domain knowledge. This model requires no signal prior, no noise model estimation, and no paired training images. The only assumption is that the inputs are acquired with identical experimental configurations. To evaluate the restoration performance of our model, as it is impossible to obtain ground truth for our experimental dataset, we propose consecutive structural similarity (CSS) for image quality assessment, based on the fact that the structures remain much the same as the previous frame(s) within small scan intervals. We demonstrate the superiority of our model by providing evaluation in terms of CSS and visual quality on different experimental datasets.

• 국제물리치료학회지
• /
• 제12권1호
• /
• pp.2261-2266
• /
• 2021

Background: Deep learning related research works on website medical images have been actively conducted in the field of health care, however, articles related to the musculoskeletal system have been introduced insufficiently, deep learning-based studies on classifying orthopedic manual therapy images would also just be entered. Objectives: To create a deep learning model that categorizes cervical mobilization images and establish a web application to find out its clinical utility. Design: Research and development. Methods: Three types of cervical mobilization images (central posteroanterior (CPA) mobilization, unilateral posteroanterior (UPA) mobilization, and anteroposterior (AP) mobilization) were obtained using functions of 'Download All Images' and a web crawler. Unnecessary images were filtered from 'Auslogics Duplicate File Finder' to obtain the final 144 data (CPA=62, UPA=46, AP=36). Training classified into 3 classes was conducted in Teachable Machine. The next procedures, the trained model source was uploaded to the web application cloud integrated development environment (https://ide.goorm.io/) and the frame was built. The trained model was tested in three environments: Teachable Machine File Upload (TMFU), Teachable Machine Webcam (TMW), and Web Service webcam (WSW). Results: In three environments (TMFU, TMW, WSW), the accuracy of CPA mobilization images was 81-96%. The accuracy of the UPA mobilization image was 43~94%, and the accuracy deviation was greater than that of CPA. The accuracy of the AP mobilization image was 65-75%, and the deviation was not large compared to the other groups. In the three environments, the average accuracy of CPA was 92%, and the accuracy of UPA and AP was similar up to 70%. Conclusion: This study suggests that training of images of orthopedic manual therapy using machine learning open software is possible, and that web applications made using this training model can be used clinically.