• The KIPS Transactions:PartB
• /
• v.14B no.5
• /
• pp.351-360
• /
• 2007

Video Completion refers to a computer vision technique which restores damaged images by filling missing pixels with suitable color in a video sequence. We propose a new video completion technique to fill in image holes which are caused by removing an unnecessary object in a video sequence, where two objects cross each other in the presence of camera motion. We remove the closer object from a camera which results in image holes. Then these holes are filled by color information of some others frames. First of all, spatio-temporal volumes of occluding and occluded objects are created according to the centroid of the objects. Secondly, a temporal search technique by voxel matching separates and removes the occluding object. Finally. these holes are filled by using spatial search technique. Seams on the boundary of completed pixels we removed by a simple blending technique. Experimental results using real video sequences show that the proposed technique produces new completed videos.

• 한국HCI학회:학술대회논문집
• /
• 2007.02a
• /
• pp.281-286
• /
• 2007

비디오 복원(video completion)은 비디오 영상에서 색상값에 대한 정보가 없는 픽셀에 적절한 색을 채워 영상을 복원하는 기술이다. 본 논문에서는 움직이는 두 물체가 교차하는 입력영상으로부터 하나의 물체를 제거함으로 발생하는 홀(hole)을 채우는 비디오 복원 기술을 제안한다. 입력 영상에서의 두 물체 중 카메라와 가까운 물체를 제거함으로써 영상의 홀이 발생하게 되고, 이 홀을 다른 프레임들의 정보를 이용하여 채움으로써 가려진 물체를 복원한다. 모든 프레임에 대해 각 물체의 중심을 추정하여 물체의 중심을 기준으로 시-공간 볼륨(spatio-temporal volume)을 생성하고, 복셀 매칭(voxel matching)을 통한 시간적 탐색을 수행한 후 두 물체를 분리한다. 가리는 물체 영역으로 판단 된 부분을 삭제하고 공간적 탐색 방법을 이용하여 홀을 채워 가려짐이 있는 물체를 복원하는 과정을 소개한다. 실험 결과를 통해 제안한 기술이 비교적 자연스러운 결과를 얻을 수 있다는 것을 보여준다.

• Journal of KIISE:Computing Practices and Letters
• /
• v.12 no.1
• /
• pp.56-65
• /
• 2006

Recently the study on service robots has been proliferated in many fields, and there are active developments for indoor services such as supporting for elderly people. It is important for robot to recognize objects and situations appropriately for effective and accurate service. Conventional object recognition methods have been based on the pre-defined geometric models, but they have limitations in indoor environments with uncertain situation such as the target objects are occluded by other ones. In this paper we propose a Bayesian network model to reason the probability of target objects for effective detection. We model the relationships between objects by activities, which are applied to non-static environments more flexibly. Overall structure is constructed by combining common-cause structures which are the units making relationship between objects, and it makes design process more efficient. We test the performance of two Bayesian networks for verifying the proposed Bayesian network model through experiments, resulting in accuracy of $86.5\%$ and $89.6\%$ respectively.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2005.11a
• /
• pp.195-198
• /
• 2005

최근 실내 환경에서 영상 정보를 사용하여 로봇이 서비스를 제공하기 위한 연구가 활발하다. 과거 영상 처리 접근 방법은 산업 환경과 같은 예측 가능한 환경을 바탕으로 미리 정의된 기하학적 모델을 통해 상황을 인식하였기에, 이를 실내 환경과 같은 가변적인 환경에 적용할 시 성능이 저하된다. 이에 지식을 기반으로 불확실성을 해결하여 정확도를 향상 시킴으로써 영상 인식 성능을 높이기 위한 다양한 연구가 진행되어 왔다. 본 논문에서는 실내에서 활동하는 서비스 로봇의 물체인식 성능을 향상시키기 위해, 대상 물체가 다른 물체에 의해서 가려져 있는 경우 대상 물체의 존재 여부를 추론하기 위한 베이지안 네트워크 모델링 방법을 제안한다. 제안하는 방법은 작은 단위로 설계된 베이지안 네트워크들을 상황에 따라 결합하여 추론 모델이 구성되게 하였고 물체간의 관계를 효과적으로 표현하고 초기 확률 값을 단일하게 유지하기 위해 제안된 확률 값 설정 방법을 사용하였다. 실험은 물체 관계를 추론하는 모듈의 성능을 검증하기 위해 수행되었는데, 5가지 장소에서 82.8$\%$의 정확도를 보여주었다.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.15 no.7
• /
• pp.817-822
• /
• 2005

Recently tile study that exploits visual information for tile services of robot in indoor environments is active. Conventional image processing approaches are based on the pre-defined geometric models, so their performances are likely to decrease when they are applied to the uncertain and dynamic environments. For this, diverse researches to manage the uncertainty based on the knowledge for improving image recognition performance have been doing. In this paper we propose a Bayesian network modeling method for predicting the existence of target objects when they are occluded by other ones for improving the object detection performance of the service robots. The proposed method makes object relationship, so that it allows to predict the target object through observed ones. For this, we define the design method for small size Bayesian networks (primitive Bayesian netqork), and allow to integrate them following to the situations. The experiments are performed for verifying the performance of constructed model, and they shows $82.8\%$ of accuracy in 5 places.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2013.10a
• /
• pp.1027-1030
• /
• 2013

For the detection of objects in the videos, there are various ways that use the frequency transformation. In the videos, the images of objects could be changed slightly. Object detection methods using frequency transformation such as ASEF and MOSSE have the ability to renew the detection filter in order to deal with the change of object images. But these approaches are likely to fail the detection because the image changes often occur when they come out again after being hidden by other objects. What is worse, when they show up again, they appear in another place, not the original one. In this paper, a new proposal is present so that the detection can be carried out efficiently even when the images come out in other place, and the failure of the detection can be reduced.