Showing papers in "Progress in Informatics in 2010"

Diminished reality using plane-sweep algorithm with weakly-calibrated cameras

Abstract: We present a plane-sweep algorithm for removing occluding objects in front of the objective scene from multiple weakly-calibrated cameras. Projective grid space (PGS), a weak cameras calibration framework, is used to obtain geometrical relations between cameras. Plane-sweep algorithm works by implicitly reconstructing the depth maps of the targeted view. By excluding the occluding objects from the volume of the sweeping planes, we can generate new views without the occluding objects. The results show the effectiveness of the proposed method and it is fast enough to run in several frames per second on a consumer PC by implementing the proposed plane-sweep algorithm in graphics processing unit (GPU).

Fast face clustering based on shot similarity for browsing video

Abstract: In this paper, we propose a new approach for clustering faces of characters in a recorded television title. The clustering results are used to catalog video clips based on subjects’ faces for quick scene access. The main goal is to obtain a result for cataloging in tolerable waiting time after the recording, which is less than 3 minutes per hour of video clips. Although conventional face recognition-based clustering methods can obtain good results, they require considerable processing time. To enable high-speed processing, we use similarities of shots where the characters appear to estimate corresponding faces instead of calculating distance between each facial feature. Two similar shot-based clustering (SSC) methods are proposed. The first method only uses SSC and the second method uses face thumbnail clustering (FTC) as well. The experiment shows that the average processing time per hour of video clips was 350 ms and 31 seconds for SSC and SSC+FTC, respectively, despite the decrease in the average number of different person’s faces in a catalog being 6.0% and 0.9% compared to face recognition-based clustering.

Recognizing multiple objects based on co-occurrence of categories

Abstract: Most previous methods for generic object recognition explicitly or implicitly assume that an image contains objects from a single category, although objects from multiple categories often appear together in an image. In this paper, we present a novel method for object recognition that explicitly deals with objects of multiple categories coexisting in an image. Furthermore, our proposed method aims to recognize objects by taking advantage of a scene’s context represented by the co-occurrence relationship between object categories. Specifically, our method estimates the mixture ratios of multiple categories in an image via MAP regression, where the likelihood is computed based on the linear combination model of frequency distributions of local features, and the prior probability is computed from the co-occurrence relation. We conducted a number of experiments using the PASCAL dataset, and obtained the results that lend support to the effectiveness of the proposed method.

Object segmentation under varying illumination: Stochastic background model considering spatial locality

Abstract: We propose a new method for background modeling. Our method is based on the two complementary approaches. One uses the probability density function (PDF) to approximate background model. The PDF is estimated non-parametrically by using Parzen density estimation. Then, foreground object is detected based on the estimated PDF. The method is based on the evaluation of the local texture at pixel-level resolution which reduces the effects of variations in lighting. Fusing those approachs realizes robust object detection under varying illumination. Several experiments show the effectiveness of our approach.

Multiple view geometry of projector-camera systems from virtual mutual projection

Abstract: Recently, projector camera systems have been used actively for image synthesis and for 3D reconstruction. For using the projector camera systems in these applications, it is very important to compute the geometry between projectors and cameras accurately. Recently, it has been shown that by using the mutual projection of cameras, multiple cameras are calibrated quite accurately. However, this property cannot be used for projector camera systems, since projectors are light-emitting devices and the projection of cameras cannot be obtained in projector images. In this paper, we show that by using the shadow of cameras and the shadow of projectors generated by projector light, we can generate virtual mutual projections between projectors and cameras, and projectors and projectors. These virtual mutual projections can be used for calibrating projector-camera systems quite accurately. Furthermore, the calibration can be achieved without using any 3D points unlike the existing calibration methods. The accuracy of the proposed method is evaluated by using real and synthetic images.

Associating faces and names in Japanese photo news articles

Abstract: We propose a system which extracts faces and person names from news articles with photos on the Web and associates them automatically. The system detects face images in news photos with a face detector and extracts person names from news text with a morphological analyzer. In addition, the bag-of-keypoints representation is applied to the extracted face images for filtering out non-face images. The system uses the eigenface representation as image features of the extracted faces, and associates them with the extracted names by the modified k-means clustering in the eigenface subspace. In the experiment, we obtained the 66% precision rate at most regarding association of faces and names.