• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.2
• /
• pp.231-237
• /
• 2020

In this paper, we provide an efficient hybrid video blending scheme to improve the naturalness of composite video in Poisson equation-based composite methods. In image blending process, various blending methods are used depending on the purpose of image composition. The hybrid blending method proposed in this paper has the characteristics that there is no seam in the composite video and the color distortion of the object is reduced by properly utilizing the advantages of Poisson blending and alpha blending. First, after blending the source object by the Poisson blending method, the color difference between the blended object and the original object is compared. If the color difference is equal to or greater than the threshold value, the object of source video is alpha blended and is added together with the Poisson blended object. Simulation results show that the proposed method has not only better naturalness than Poisson blending and alpha blending, but also requires a relatively small amount of computation.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.22 no.4
• /
• pp.617-623
• /
• 2018

In this paper, we propose a novel seamless video composition method based on shape matching and Poisson equation. Video composition method consists of video segmentation process and video blending process. In the video segmentation process, the user first sets a trimap for the first frame, and then performs a grab-cut algorithm. Next, considering that the performance of video segmentation may be reduced if the color, brightness and texture of the object and the background are similar, the object region segmented in the current frame is corrected through shape matching between the objects of the current frame and the previous frame. In the video blending process, the object of source video and the background of target video are blended seamlessly using Poisson equation, and the object is located according to the movement path set by the user. Simulation results show that the proposed method has better performance not only in the naturalness of the composite video but also in computational time.

• Journal of the Korea Computer Graphics Society
• /
• v.20 no.3
• /
• pp.1-8
• /
• 2014

We propose a method to compose a foreground object into a background image, where the foreground object is a part (or a region) of an image taken by a front-facing camera and the background image is a whole image taken by a back-facing camera in a smart phone at the same time. Recent high-end cell-phones have two cameras and provide users with preview video before taking photos. We extract the foreground object that is moving along with the front-facing camera using the optical flow during the preview. We compose the extracted foreground object into a background image using a simple image composition technique. For better-looking result in the composed image, we apply a border smoothing technique using a weighted-border mask to blend transparency from background to foreground. Since constructing and grouping pixel-level dense optical flow are quite slow even in high-end cell-phones, we compute a mask to extract the foreground object in low-resolution image, which reduces the computational cost greatly. Experimental result shows the effectiveness of our extraction and composition techniques, with much less computational time in extracting the foreground object and better composition quality compared with Poisson image editing technique which is widely used in image composition. The proposed method can improve limitedly the color bleeding artifacts observed in Poisson image editing using weighted-border blending.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.27 no.1
• /
• pp.8-14
• /
• 2023

The existing self-supervised learning-based CutPaste generated synthetic data by cutting and attaching specific patches from normal images and then performed anomaly detection. However, this method has a problem in that there is a clear difference in the boundary of the patch. NSA for solving these problems have achieved higher anomaly detection performance by generating natural synthetic data through Poisson Blending. However, NSA has the disadvantage of having many hyperparameters that need to be adjusted for each class. In this paper, synthetic data similar to normal were generated by a simple method of making the size of the synthetic patch very small. At this time, since the patches are so locally synthesized, models that learn local features can easily overfit synthetic data. Therefore, we performed anomaly detection using gMLP, which learns global features, and even with simple synthesis methods, we were able to achieve higher performance than conventional self-supervised learning techniques.

• Proceedings of the Korea Multimedia Society Conference
• /
• 2012.05a
• /
• pp.262-263
• /
• 2012

다중 센서의 영상, 예를 들어 가시광 영상과 적외선 영상은 서로 다른 특징을 가지고 있기 때문에 본 논문에서는 IR 영상의 특징을 보존한 새로운 혼합기법을 제안하다. 이러한 혼합기법은 의료 영상, 보안 영상 등에서 매우 중요하고 다양하게 다루어진다. 일반적인 혼합기법을 사용하게 되면 영상간의 특색 때문에 혼합 시 조화롭지 못하는 문제점을 가진다. 이러한 문제점을 해결하기 위해서 본 논문에서는 중요도 맵을 추출하고 그 영역에 대하여 포아송 블랜딩을 통해 두 개의 다른 특징을 가시광 영상을 혼합한다. 제안한 알고리즘은 기존의 연구와 다르게 혼합할 영역을 수동으로 지정하는 것이 아니라 자동적으로 추출하고, 가시광 영상에 IR 영상에서만 검출되는 영역을 결합한 새로운 결과를 얻을 수 있었다.