• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.10a
• /
• pp.255-262
• /
• 1998

This paper presents a translation, rotation and scale invariant methodology for the recognition of closed planar shape images using the bispectrum of a contour sequence and the learning vector quantization(LVQ) neural classifier. The contour sequences obtained from the closed planar images represent the Euclidean distance between the centroid and all boundary pixels of the shape, and are related to the overall shape of the images. The higher order spectra based on third order cumulants is applied to tihs contour sample to extract fifteen bispectral feature vectors for each planar image. There feature vector, which are invariant to shape translation, rotation and scale transformation, can be used to represent two0dimensional planar images and are fed into a neural network classifier. The LVQ architecture is chosen as a neural classifier because the network is easy and fast to train, the structure is relatively simple. The experimental recognition processes with eight different hapes of aircraft images are presented to illustrate the high performance of this proposed method even the target images are significantly corrupted by noise.

• Investigative Magnetic Resonance Imaging
• /
• v.1 no.1
• /
• pp.142-147
• /
• 1997

Purpose: To compare the effectiveness of the in-phase (IP) sequence and the opposed-phase (Op) sequence in the detection of focal hepatic lesions in the single breath-hold hepatic MR imaging with fast gradient T1-weighted pulse sequences. Materials and Methods: IP and OP T1-weighted breath-hold imaging was performed using fast gradient echo sequences in 45 patients referred for known focal hepatic lesions, in which 78 lesions were detected. Three blind readers independently reviewed the images for lesion detectability. The signal-to-noise ratio (SNR) of the liver, the lesion-to-liver contrast-to-noise ratio (CNR) and the liver-to-spleen CNR were also compared. A consensus was reached by three readers to determine which sequence is better in image quality. Results: On OP images, 61(78%), 61(78%), and 63(89%) lesions were correctly identified for reader 1, 2 and 3, respectively. On IP images, 66(85%), 65(83%), and 65(93%) lesions were detected for each reader, respectively. When two image sets were combined, 71(91 %), 69(88 %), and 76(97%) lesions respectively were detected for each reader. In cases of hepatocellular carcinoma, liver-to-Iesion CNR was greater on the OP images(p (0.05), but in other lesions significant difference was not demonstrated. Liver-to-spleen CNR was higher on OP images(p ( 0.1), but the SNR of the liver was higher on the IP images. Conclusion: Use of both IP and OP imaging can be helpful to avoid erroneous missing of some focal hepatic lesions.

• The Bulletin of The Korean Astronomical Society
• /
• v.44 no.2
• /
• pp.44.1-44.1
• /
• 2019

We generate new intermediate images between observed consecutive solar images using NVIDIA's SuperSloMo that is a novel video interpolation method. This technique creates intermediate frames between two successive frames to form a coherent video sequence for both spatially and temporally. By using SuperSloMo, we create 600 images (12-second interval) using the observed 121 SDO/AIA 304 Å images (1-minute interval) of a filament eruption event on December 3, 2012. We compare the generated images with the original 12-second images. For the generated 480 images the correlation coefficient (CC), the relative error (R1), and the normalized mean square error (R2) are 0.99, 0.40, and 0.86, respectively. We construct a video made of the generated images and find a smoother erupting movement. In addition, we generate nonexistent 2.4-second interval images using the original 12-second interval images, showing slow motions in the eruption. We will discuss possible applications of this method.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.40 no.2
• /
• pp.29-36
• /
• 2003

This paper proposes a new 3D image transmission technique using digital watermarking for compatibility with conventional 2D digital TV. To transmit 3D images effectively, generally, we use image sequence transmission using temporal-spatial redundancy between stereo images. It is difficult for users with conventional digital TV to watch the transmitted 3D image sequence because of effectivity of 3D image compression. To improve such problem, in this paper, we perceive the concealment of new information of digital watermarking and conceal information of the other stereo image into three channels of the reference image. So we can watch the image with 3D TV as well as conventional digital TV when it is decoded.

• Journal of KIISE:Software and Applications
• /
• v.32 no.5
• /
• pp.357-365
• /
• 2005

We present a B-spline fitting method based on genetic algorithm for the extraction of object contours from the complex image sequence, where objects with similar shape and intensity are adjacent each other. The proposed algorithm solves common malfitting problem of the existing B-spline fitting methods including snakes. Classical snake algorithms have not been successful in such an image sequence due to the difficulty in initialization and existence of multiple extrema. We propose a B-spline fitting method using a genetic algorithm with a new initial population generation and fitting function, that are designed to take advantage of the contour of the previous slice. The test results show that the proposed method extracts contour of individual object successfully from the complex image sequence. We validate the algorithm by false-positive/negative errors and relative amounts of agreements.

• Proceedings of the KOSOMBE Conference
• /
• v.1990 no.11
• /
• pp.33-37
• /
• 1990

Endoscope system is the device that observe interior the body. It has some demerits because it use of optical fibers. The resolution of images depends on the number of optical fibers, it is impossible that several people observe at the same time and it needs special camera in order to record images. In order to overcome these demerits, electro endoscope system using CCD(Charge Coupled Device) has been developed recently. If the diameter of insertion tube is thinner, it is easier to be examined by thin endoscope system. At the present time, we develope these type of electro endoscope system in the department of SNU biomedical engineering. Moreover, we are researching an algorithm of the 3D-reconstruction of interior of the gaster. In this paper, a method for 3D information detection using electroendoscopic image sequence has been presented.

• 제어로봇시스템학회:학술대회논문집
• /
• 1988.10b
• /
• pp.832-836
• /
• 1988

"Motion Stereo" is quite useful for visual guidance of the robot, but most range finding algorithms of motion stereo have suffered from poor accuracy due to the quantization noise and measurement error. In this paper, 3-D position estimation and refinement scheme is proposed, and its performance is discussed. The main concept of the approach is to consider the entire frame sequence at the same time rather than to consider the sequence as a pair of images. The experiments using real images have been performed under following conditions : hand-held camera, static object. The result demonstrate that the proposed nonlinear least-square estimation scheme provides reliable and fairly accurate 3-D position information for vision-based position control of robot. of robot.

• Journal of Institute of Control, Robotics and Systems
• /
• v.17 no.10
• /
• pp.983-988
• /
• 2011

This paper presents a robust method for ground plane detection in vision-based applications based on a monocular sequence of images with a non-stationary camera. The proposed method, which is based on the reliable estimation of the homography between two frames taken from the sequence, aims at designing a practical system to detect road surface from traffic scenes. The homography is computed using a feature matching approach, which often gives rise to inaccurate matches or undesirable matches from out of the ground plane. Hence, the proposed homography estimation minimizes the effects from erroneous feature matching by the evaluation of the difference between the predicted and the observed matrices. The method is successfully demonstrated for the detection of road surface performed on experiments to fill an information void area taken place from geometric transformation applied to captured images by an in-vehicle camera system.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2022.06a
• /
• pp.182-183
• /
• 2022

The large data volume of light field (LF) image has motivated much research on how to compress the data volume more efficiently. One of the approaches is to compress LF images after representing them in the form of pseudo video sequence. In this way, the pseudo temporal redundancy between views can be exploited by motion estimation and compensation. Based on our observation that images obtained by LF cameras have small range of disparity values between adjacent views, we propose to limit the motion search range to reduce the time complexity of motion estimation. Our experimental results show that a smaller motion search range reduces the encoding time while not affecting the bitrate of H.266/VVC much.

• Progress in Medical Physics
• /
• v.10 no.1
• /
• pp.33-40
• /
• 1999

Purpose: To simulate and measure the signal intensity of various tissues near bone interface in 2D and 3D neurological MR images. Materials and Methods: In neurological proton density (PD) weighted images, every component in the head including cerebrospinal fluid (CSF), muscle and scalp, with the exception of bone, are visualised. It is possible to acquire images in 2D or 3D. A 2D fast spin-echo (FSE) sequence is chosen for the 2D acquisition and a 3D gradient-echo (GE) sequence is chosen for the 3D acquisition. To find out the signal intensities of CSF, muscle and fat (or scalp) for the 2D spin-echo(SE) and 3D gradient-echo (GE) imaging sequences, the theoretical signal intensities for 2D SE and 3D GE were calculated. For the 2D fast spin-echo (FSE) sequence, to produce the PD weighted image, long TR (4000 ms) and short TE$_{eff}$ (22 ms) were employed. For the 3D GE sequence, low flip angle (8$^{\circ}$) with short TR (35 ms) and short TE (3 ms) was used to produce the PD weighted contrast. Results: The 2D FSE sequence has CSF, muscle and scalp with superior image contrast and SNR of 39 - 57 while the 3D GE sequence has CSF, muscle and scalp with broadly similar image contrast and SNR of 26 - 33. SNR in the FSE image were better than those in the GE image and the skull edges appeared very clearly in the FSE image due to the edge enhancement effect in the FSE sequence. Furthermore, the contrast between CSF, muscle and scalp in the 2D FSE image was significantly better than in the 3D GE image, due to the strong signal intensities (or SNR) from CSF, muscle and scalp and enhanced edges of CSF. Conclusion: The signal intensity of various tissues near bone interface in neurological MR images has been simulated and measured. Both the simulation and imaging of the 2D SE and 3D GE sequences have CSF, fat and muscle with broadly similar image intensity and SNR's and have succeeded in getting all tissues about the same signal. However, in the 2D FSE sequence, image contrast between CSF, muscle and scalp was good and SNR was relatively high, imaging time was relatively short.