• 한국HCI학회:학술대회논문집
• /
• 2009.02a
• /
• pp.625-630
• /
• 2009

Researches of image-based 3D reconstruction have recently produced a number of good results, but they assumed that the accurate foreground to be reconstructed is already extracted from each input image. This paper proposes a novel approach to extract more accurate foregrounds by iteratively performing foreground extraction and 3D reconstruction in a manner similar to an EM algorithm on regions segmented in an initial stage, called segments. Here, the segments should preserve foreground boundaries to compensate for the boundary errors generated by visual hull, simple 3D reconstruction to minimize the computational time, and should also be composed of the small number of sets to minimize the user input. Therefore, we utilize image segmentation using the graph-cuts method, which minimizes energy function composed of data and smoothness terms, and the two methods are iteratively performed until the energy function is optimized. In the experiments, more accurate results of the foreground, especially in boundaries, were obtained, although the proposed method used a simple 3D reconstruction method.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.10 no.1
• /
• pp.127-135
• /
• 2006

The image-based 3D modeling is the technique of generating a 3D graphic model from images acquired using cameras. It is being researched as an alternative technique for the expensive 3D scanner. In this paper, I propose the image-based 3D modeling system using calibrated camera. The proposed algorithm for rendering 3D model is consisted of three steps, camera calibration, 3D shape reconstruction and 3D surface generation step. In the camera calibration step, I estimate the camera matrix for the image aquisition camera. In the 3D shape reconstruction step, I calculate 3D volume data from silhouette using pyramidal volume intersection. In the 3D surface generation step, the reconstructed volume data is converted to 3D mesh surface. As shown the result, I generated relatively accurate 3D model.

• The KIPS Transactions:PartB
• /
• v.17B no.3
• /
• pp.183-188
• /
• 2010

A factorization-based 3D reconstruction system is realized to recover 3D scene from an image sequence. The image sequence is captured from uncalibrated perspective camera from several views. Many matched feature points over all images are obtained by feature tracking method. Then, these data are supplied to the 3D reconstruction module to obtain the projective reconstruction. Projective reconstruction is converted to Euclidean reconstruction by enforcing several metric constraints. After many triangular meshes are obtained, realistic reconstruction of 3D models are finished by texture mapping. The developed system is implemented in C++, and Qt library is used to implement the system user interface. OpenGL graphics library is used to realize the texture mapping routine and the model visualization program. Experimental results using synthetic and real image data are included to demonstrate the effectiveness of the developed system.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2021.06a
• /
• pp.327-330
• /
• 2021

본 논문은 RGB 영상 데이터셋의 일부만을 지도학습하여(Sparsely-supervised learning) Annotation 되지 않은 영상에 대해 손-객체의 3D 포즈를 복원하기 위한 방법을 제안한다. 기존의 연구에서는 손-객체의 포즈에 해당하는 6DoF 만을 학습 데이터로 활용한다. 이와 달리, 본 논문에서는 정확도 향상을 위해 복원된 결과를 동일한 입력 영상 내에서 비교 가능하도록 3D 모델로 복원한 결과를 입력 영상의 마스크로 만들어 학습에 반영하였다. 구체적으로 추정된 포즈로 만들어낸 마스크를 입력 영상에 적용한 결과와 Ground-truth 포즈를 적용한 영상을 학습 시에 손실 함수에 반영하였다. 비교 실험을 통해 제안된 방법이 해당 방법을 적용하지 않은 경우 보다 3D 매쉬 오차가 적었음을 확인할 수 있었다.

• Journal of Broadcast Engineering
• /
• v.24 no.2
• /
• pp.264-272
• /
• 2019

In this paper, we propose a 3D human body reconstruction and refinement method from the frames extracted from a video to obtain natural and smooth motion in temporal domain. Individual frames extracted from the video are fed into convolutional neural network to estimate the location of the joint and the silhouette of the human body. This is done by projecting the parameter-based 3D deformable model to 2D image and by estimating the value of the optimal parameters. If the reconstruction process for each frame is performed independently, temporal consistency of human pose and shape cannot be guaranteed, yielding an inaccurate result. To alleviate this problem, the proposed method analyzes and interpolates the principal component parameters of the 3D morphable model reconstructed from each individual frame. Experimental result shows that the erroneous frames are corrected and refined by utilizing the relation between the previous and the next frames to obtain the improved 3D human reconstruction result.

• Journal of Internet of Things and Convergence
• /
• v.9 no.1
• /
• pp.89-95
• /
• 2023

In January 2007, the Remains Excavation and Investigation Team of the Ministry of National Defense was established, and full-scale excavation of remains was promoted. Currently, the scope of the excavation is being expanded to Baekma Hill within the DMZ, where fierce battles were fought during the Korean War. Now, many remains and remains are being excavated in Baekma hill. Most are in damaged condition. Therefore, in this study, the original form of the excavated remains was restored using 3D scanning and 3D modeling. This digital restoration method can be an alternative to compensate for the disadvantages of the manual method. Currently, various digital restorations using 3D technology are active in the field of cultural heritage. Digitally restored materials can be used as basic data for digital heritage. Based on this, various contents related to excavation of remains and patriots and veterans can be developed. Furthermore, if digital human restoration is made based on the excavated remains, it will be possible to reproduce the appearance of the dead.

• Journal of the Korea Computer Industry Society
• /
• v.2 no.3
• /
• pp.297-308
• /
• 2001

We propose an algorithm which can reconstruct the 3D information from geographical information. The conventional techniques, the triangular patches and the Random Fractal Midpoint Displacement (RFMD) method, etc., have often been used to reconstruct natural images. While the RFMD method using Gaussian distribution obtains good results for the symmetric images, it is not reliable on asymmetric images immanent in the nature. Our proposed algorithm employs neural networks for the RFMD method to present the asymmetrical images. By using a neural network for reconstructing the 3D images, we can utilize statistical characteristics of irregular data. We show that our algorithm has a better performance than others by the point of view on the similarity evaluation. And, it seems that our method is more efficient for the mountainous topography which is more rough and irregular.

• Conservation Science in Museum
• /
• v.22
• /
• pp.85-96
• /
• 2019

The restoration of ceramics excavated in fragments is limited by the difficulty of inferring the overall shape of the original object. However, recent innovations in digital technology can help to overcome the limits of conventional restoration using handwork. This study explored the potential of digital technology by digitally restoring a shard from a white porcelain water dropper excavated at a kiln site at Sindae-ri. In order to complete the digital restoration, 3D scanning was applied to obtain scan data, and 3D modeling and texture mapping were performed. In this way, three-dimensional data with patterns and color information was acquired and the original form of the water dropper could be ascertained based on the shard. The study found that the data acquired from digital restoration can be used for various purposes, including for obtaining data on cross-sections or missing portions of a relic.

• Journal of KIISE:Computer Systems and Theory
• /
• v.30 no.3_4
• /
• pp.168-185
• /
• 2003

This paper presents a progressive algorithm that not only can narrow down the search domain in the course of face identification but also can fast reconstruct various 3D objects from a sketch drawing. The sketch drawing, edge-vertex graph without hidden line removal, which serves as input for reconstruction process, is obtained from an inaccurate freehand sketch of a 3D wireframe object. The algorithm is executed in two stages. In the face identification stage, we generate and classify potential faces into implausible, basis, and minimal faces by using geometrical and topological constraints to reduce search space. The proposed algorithm searches the space of minimal faces only to identify actual faces of an object fast. In the object reconstruction stage, we progressively calculate a 3D structure by optimizing the coordinates of vertices of an object according to the sketch order of faces. The progressive method reconstructs the most plausible 3D object quickly by applying 3D constraints that are derived from the relationship between the object and the sketch drawing in the optimization process. Furthermore, it allows the designer to change viewpoint during sketching. The progressive reconstruction algorithm is discussed, and examples from a working implementation are given.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2010.07a
• /
• pp.415-416
• /
• 2010

이 논문에서 우리는 텍스쳐 정보가 적은 영역이 존재하는 입력 영상들에 대해서도 안정적인 복원을 도출하는 새로운 다시 점 스테레오 방법을 제시한다. 제안된 방법에서는 입력 영상들을 인접한 픽셀간의 색 유사성을 이용하여 세그먼테이션한 후, 세그먼트 단위로 다시점 스테레오를 수행한다. 특히 그 과정에서 한 영상 내의 이웃한 세그먼트들의 깊이 값 유사성, 그리고 서로 다른 시점에서 상응하는 세그먼트 간의 깊이 값 일관성을 가정하여, 텍스쳐 정보가 적은 영역에 대해서도 안정적으로 3D 점들을 생성해준다. 생성된 3D 점들은 그래프 컷 기반의 복원 알고리즘을 통해 일관된 3D 표면으로 복원 되었고, 복원 결과는 제안된 방법이 실제로 기존의 다른 다시점 스테레오에 비해 보다 안정적으로 깊이 정보를 추출할 수 있음을 보여준다. 결과적으로 제안된 방법은 보다 일반적이고 실생활에 가까운 입력 영상들에 대해서도 3D 복원을 수행할 수 있는 방향을 제시한다.