This paper surveys the results of various studies on 3-D image coding. Themes are focused on efficient compression and display-independent representation of 3-D images. Most of the works on 3-D image coding have been concentrated on the compression methods tuned for each of the 3-D image formats (stereo pairs, multi-view images, volumetric images, holograms and so on). For the compression of stereo images, several techniques concerned with the concept of disparity compensation have been developed. For the compression of multi-view images, the concepts of disparity compensation and epipolar plane image (EPI) are the efficient ways of exploiting redundancies between multiple views. These techniques, however, heavily depend on the limited camera configurations. In order to consider many other multi-view configurations and other types of 3-D images comprehensively, more general platform for the 3-D image representation is introduced, aiming to outgrow the framework of 3-D "image" communication and to open up a novel field of technology, which should be called the "spatial" communication. Especially, the light ray based method has a wide range of application, including efficient transmission of the physical world, as well as integration of the virtual and physical worlds.

We propose a new stereo image retargeting method based on the framework of shift-map image editing. Retargeting is the process of changing the image size according to the target display while preserving as much of the richness of the image as possible, and is often applied to monocular images and videos. Retargeting stereo images poses a new challenge because pixel correspondences between the stereo pair should be preserved to keep the scene's structure. The main contribution of this paper is integrating a stereo correspondence constraint into the retargeting process. Among several retargeting methods, we adopt shift-map image editing because this framework can be extended naturally to stereo images, as we show in this paper. We confirmed the effectiveness of our method through experiments.

A user wearing a Head-Mounted Display (HMD) is likely to feel isolated when sharing virtual reality (VR) experiences with Non-HMD users in the same physical space due to not being able to see the real space outside the virtual world. This research proposes a method for an HMD user to recognize the Non-HMD users' gaze and attention via a projector attached to the HMD. In the proposed approach, the projected HMD user's view is filtered darker than default, and when Non-HMD users point controllers towards the projected view, the filter is removed from a circular area for both HMD and Non-HMD users indicating which region the Non-HMD users are viewing. We conducted two user studies showing that the Non-HMD users' gaze can be recognized with the proposed method, and investigated the preferred range for the alpha value and the size of the area for removing the filter for the HMD user.

Previous studies on anomaly detection in videos have trained detectors in which reconstruction and prediction tasks are performed on normal data so that frames on which their task performance is low will be detected as anomalies during testing. This paper proposes a new approach that involves sorting video clips, by using a generative network structure. Our approach learns spatial contexts from appearances and temporal contexts from the order relationship of the frames. Experiments were conducted on four datasets, and we categorized the anomalous sequences by appearance and motion. Evaluations were conducted not only on each total dataset but also on each of the categories. Our method improved detection performance on both anomalies with different appearance and different motion from normality. Moreover, combining our approach with a prediction method produced improvements in precision at a high recall.

We present a real-time video-based rendering system using a network camera array. Our system consists of 64 commodity network cameras that are connected to a single PC through a gigabit Ethernet. To render a high-quality novel view, our system estimates a view-dependent per-pixel depth map in real time by using a layered representation. The rendering algorithm is fully implemented on the GPU, which allows our system to efficiently perform capturing and rendering processes as a pipeline by using the CPU and GPU independently. Using QVGA input video resolution, our system renders a free-viewpoint video at up to 30 frames per second, depending on the output video resolution and the number of depth layers. Experimental results show high-quality images synthesized from various scenes.

This paper discusses recognition up to intensities of mix of primary facial expressions in real time. The proposed recognition method is compatible with the MPEG-4 high level expression Facial Animation Parameter (FAP). In our method, the whole facial image is considered as a single pattern without any block segmentation. As model features, an expression vector, viz. low global frequency coefficient (DCT) changes relative to neutral facial image of a person is used. These features are robust and good enough to deal with real time processing. To construct a person specific model, apex images of primary facial expression categories are utilized as references. Personal facial expression space (PFES) is constructed by using multidimensional scaling. PFES with its generalization capability maps an unknown input image relative to known reference images. As PFES possesses linear mapping characteristics, MPEG-4 high level expression FAP can be easily calculated by the location of the input face on PFES. Also, temporal variations of facial expressions can be seen on PFES as trajectories. Experimental results are shown to demonstrate the effectiveness of the proposed method.