• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.42 no.4 s.304
• /
• pp.87-94
• /
• 2005

3D image mosaicking is useful for 3D visualization of the roadside scene of urban area by projecting 2D images to the 3D planes. When a sequence of images are filmed from a side-looking video camera passing long distance areas, the ghost effect in which same objects appear repeatively occurs. To suppress such ghost effect, the long distance range areas are detected by using the distance between the image frame and the 3D coordinate of tracked optical flows. The ghost effects are suppressed by projecting the part of image frames onto 3D multiple planes utilizing vectors passing the focal point of frames and a virtual focal point. The virtual focal point is calculated by utilizing the first and last frames of the long distance range areas. We demonstrate algorithm that creates efficient 3D Panoramic mosaics without the ghost effect at the long distance area.

• Progress in Medical Physics
• /
• v.24 no.4
• /
• pp.243-252
• /
• 2013

The aim of this study was to clarify the impacts of acquisition parameters on artifacts in four-dimensional computed tomography (4D CT) images, such as the partial volume effect (PVE), partial projection effect (PPE), and mis-matching of initial motion phases between adjacent beds (MMimph) in cine mode scanning. A thoracic phantom and two cylindrical phantoms (2 cm diameter and heights of 0.5 cm for No.1 and 10 cm for No.2) were scanned using 4D CT. For the thoracic phantom, acquisition was started automatically in the first scan with 5 sec and 8 sec of gantry rotation, thereby allowing a different phase at the initial projection of each bed. In the second scan, the initial projection at each bed was manually synchronized with the inhalation phase to minimize the MMimph. The third scan was intentionally un-synchronized with the inhalation phase. In the cylindrical phantom scan, one bed (2 cm) and three beds (6 cm) were used for 2 and 6 sec motion periods. Measured target volume to true volume ratios (MsTrueV) were computed. The relationships among MMimph, MsTrueV, and velocity were investigated. In the thoracic phantom, shorter gantry rotation provided more precise volume and was highly correlated with velocity when MMimph was minimal. MMimph reduced the correlation. For moving cylinder No. 1, MsTrueV was correlated with velocity, but the larger MMimph for 2 sec of motion removed the correlation. The volume of No. 2 was similar to the static volume due to the small PVE, PPE, and MMimph. Smaller target velocity and faster gantry rotation resulted in a more accurate volume description. The MMimph was the main parameter weakening the correlation between MsTrueV and velocity. Without reducing the MMimph, controlling target velocity and gantry rotation will not guarantee accurate image presentation given current 4D CT technology.

• Journal of Digital Convergence
• /
• v.10 no.4
• /
• pp.191-197
• /
• 2012

In this paper we present creating 3D image from 2D image by extract initial depth values calculated from focal values. The initial depth values are created by using the extracted focal information, which is calculated by the comparison of original image and Gaussian filtered image. This initial depth information is allocated to the object segments obtained from normalized cut technique. Then the depth of the objects are corrected to the average of depth values in the objects so that the single object can have the same depth. The generated depth is used to convert to 3D image using DIBR(Depth Image Based Rendering) and the generated 3D image is compared to the images generated by other techniques.

• Journal of the Institute of Convergence Signal Processing
• /
• v.6 no.4
• /
• pp.176-181
• /
• 2005

This paper presents an equal bit rate control algorithm utilizing the statistical change between the previous frame and the current frame. The previous studies on the model-based rate control have focused on the models of bit rate and distortion in types of coders, in terms of the quantization parameter. The proposed algorithm improves the typical model-based rate control by updating a model parameter instead of modeling a better model of the rate and distortion. The proposed algorithm updates this model parameter by recognizing the change in statistics between the previous frame and the current frame. We implement the proposed algorithm in MPEG-4 coders and verify its performance while comparing it to the TMN8's approach (up to 0.6dB of improvement).

• Journal of the Institute of Electronics Engineers of Korea TE
• /
• v.42 no.4
• /
• pp.21-26
• /
• 2005

This paper presents a watermarking method using similarity between frames in the scene to resist to various attacks and to improve the video quality. This method inserts and detects a watermark in the frame pair every 3 frame. The experimental simulations show that the video quality is improved more than 45dB compared with previous methods and the watermark is resistant to frame drop, MPEG compression and low pass filter attacks.

• Proceedings of the Korean Information Science Society Conference
• /
• 2003.04a
• /
• pp.629-631
• /
• 2003

대부분의 상용 CIS 소프트웨어에서는 시공간분석 결과를 2 차원상에 가시화 하거나 해당 분석결과를 단순히 출력해 주는 형태이다. 3차원상의 질의 볼륨에 대한 3차원 공간분석 또한 이론 및 실험적으로 많은 연구가 이루어져 왔으나, 3 차원상의 공간분석 결과 가시화 기능을 통합한 시스템은 아직까지 미흡한 실정이다. 본 논문에서는 3 차원상의 공간연산 및 3 차원에 시간속성을 부가한 4 차원 시공간 연산을 처리할 있는 4 차원 CIS 시스템을 소개한다. 제안된 기법은 2차원 수치지도와 3 차원 부가속성을 가진 데이터베이스를 이용하여 2 차원, 3 차원, 미디어 데이터간의 상호연계 및 통합된 가시화 프레임워크를 제공한다. 또한 가상세계 매핑기법을 이용하여 수작업에 의존한 별도의 이미지 프로세싱 과정 없이도 비디오상의 특정 프레임의 건물에 대한 속성질의 및 2D/3D 간의 상호 연계 기능을 제공할 수 있다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.4
• /
• pp.501-507
• /
• 2020

This paper proposes a new frame synchronization technique based on convolutional neural network (CNN). The conventional frame synchronizers usually find the matching instance through correlation between the received signal and the preamble. The proposed method converts the 1-dimensional correlator ouput into a 2-dimensional matrix. The 2-dimensional matrix is input to a convolutional neural network, and the convolutional neural network finds the frame arrival time. Specifically, in additive white gaussian noise (AWGN) environments, the received signals are generated with random arrival times and they are used for training data of the CNN. Through computer simulation, the false detection probabilities in various signal-to-noise ratios are investigated and compared between the proposed CNN-based technique and the conventional one. According to the results, the proposed technique shows 2dB better performance than the conventional method.

• Journal of Korea Multimedia Society
• /
• v.9 no.4
• /
• pp.421-433
• /
• 2006

This paper proposes a method to insert virtual objects into a real video stream based on feature tracking and camera pose estimation from a set of single-camera video frames. To insert or modify 3D shapes to target video frames, the transformation from the 3D objects to the projection of the objects onto the video frames should be revealed. It is shown that, without a camera calibration process, the 3D reconstruction is possible using multiple images from a single camera under the fixed internal camera parameters. The proposed approach is based on the simplification of the camera matrix of intrinsic parameters and the use of projective geometry. The method is particularly useful for augmented reality applications to insert or modify models to a real video stream. The proposed method is based on a linear parameter estimation approach for the auto-calibration step and it enhances the stability and reduces the execution time. Several experimental results are presented on real-world video streams, demonstrating the usefulness of our method for the augmented reality applications.

• The KIPS Transactions:PartD
• /
• v.10D no.4
• /
• pp.671-678
• /
• 2003

A digital video is usually very long temporally. requiring large storage capacity. Therefore, users want to watch pre-summarized video before they watch a large long video. Especially in the field of sports video, they want to watch a highlight video. Consequently, highlight video is used that the viewers decide whether it is valuable for them to watch the video or not. This paper proposes how to create soccer video highlight using the structural features of the caption such as temporal and spatial features. Caption frame intervals and caption key frames are extracted by using those structural features. And then, highlight video is created by using scene relocation, logical indexing and highlight creation rule. Finally. retrieval and browsing of highlight and video segment is performed by selection of item on browser.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.40 no.1
• /
• pp.72-78
• /
• 2003

A new self-organizing neuro-fuzzy network based rate control algorithm for MPEG-4 video encoder is proposed in this paper. Contrary to the traditional methods that construct the rate-distorion (RD) model based on experimental equations, the proposed method effectively exploits the non-stationary property of the video date with neuro-fuzzy network that self-organizes the RD model online and adaptively updates the structure. The method needs not require off-line pre-training; hence it is geared toward real-time coding. The comparative results through the experiments suggest that our proposed rate control scheme encodes the video sequences with less frame skip, providing good temporal quality and higher PSNR, compared to VM18.0.