We propose an algorithm for motion extraction from Time-Varying Mesh (TVM). TVM is a sequence of 3D models made for real-world dynamic objects. TVM store detailed information about 3D objects such as their shape, color, and motion. Therefore, TVM is an attractive technology for next-generation multimedia. However, the management of a TVM database, which is used for archiving and retrieving, is a significant problem because it is difficult to locate and track feature points in TVM. For efficient and effective archive and retrieval systems, motion extraction and tracking is essential. In our method, skeletons are extracted from TVM using a Reeb graph, and motion tracking is achieved by fitting a reference skeleton to the other skeletons. We developed a geodesic function using principal component analysis (PCA) for fast and noiseless skeleton extraction. Robust fitting was achieved by implementing an end-node tracking strategy. In addition, we have developed an efficient motion retrieval system compatible with conventional motion capture data using the extracted skeletons. As a result, we have developed a low-cost query generation technique using motion-capture systems and cross search between TVM and motion capture data.

View full abstract

We developed a method of interactive virtual camerawork transition control based on a hidden Markov model with camera orientation as a hidden state and camera distance as an observable output. This method enables a novice user to interactively create a 3-D CG animation by suggesting appropriate camerawork based on the grammar of the film language. We conducted an experimental evaluation in which 363 novice users created 3-D CG animations with the digital movie director system using this method. The results showed a 30.7 % improvement in camera setting efficiency. Time spent writing dialogues increased by 14.7 % although the overall time taken to create the animation was shortened by 4.7 %, suggesting more time was available for creative thinking. Moreover, the ratio of shots that violated the grammar decreased from 9.9 to 4.8 %.

View full abstract

We propose a pre-coded OFDM iterative detection scheme. The proposed scheme recovers the pre-coded data symbols that correspond to the carriers whose reception powers are degraded by frequency-selective fading by using the data symbols after decision, and the data symbols are then detected again using the recovered pre-coded data symbols.We confirmed by numerical analysis that better performance under severe frequency-selective fading channels can be achieved by iteratively repeating the proposed scheme.

View full abstract

We propose a method for precise overlapping of projected images using multiple steerable projectors. When steerable projectors are controlled simultaneously, two problems occur: (1) because the flatness of the projected planes is not suited to the real environment, each projected image from each multiple projector that is calibrated with just a part of the plane cannot always be located to the same position as that of an image from another projector, even when the images are intended to be located at the same position; and (2) because the projectors run asynchronously it is impossible to display a moving image by precisely tiling or overlaying the multiple projected images. To overcome this (1), we propose a method that preliminarily measures the misalignments through every plane in the environment, and hence displays the image without misalignment. For (2), we describe a method for the synchronization of moving and scaling between multiple projections.

View full abstract

We have developed a method that can remove view-disturbing noise from image sequences by spatio-temporal image processing. In outdoor environments on rainy days, pictures taken by a camera are often degraded because of adherent noise, such as water drops on the surface of the lens-protecting glass of the camera. To solve this problem, our method uses image sequences captured with a pan-tilt camera. The method uses a spatio-temporal image to extract adherent noise regions by examining trajectory differences between adherent noise and other objects in cross-section images. Finally, adherent noise regions are filled in with surrounding information in the cross-section image. Experimental results showed our method is effective.

View full abstract

This paper discusses a frame rate conversion (FRC) technology that improves LCD picture quality (PQ) by doubling the frame rate to 120 Hz; an observer study on the resulting quality is also reported. To achieve a high-performance LCD-TV system, we first re-consider fundamental design issues, such as characteristics of the human visual system, motion content in real broadcast video, and current viewing conditions related to screen size. We then discuss how FRC is effective in improving the PQ of large-screen LCD-TVs; we show that onscreen captions can be particularly improved. Next, we discuss an FRC technology using a novel approach that improves the PQ of onscreen captions; it also reduces motion vector estimation problems. A subjective study using paired comparisons of two LC-TVs with this FRC technology showed that at motion speeds below 21 deg/sec (80% of the onscreen captions), the PQ could be increased to above that of an acceptance limit of 3.5.

View full abstract

We have developed a fast shading and shadowing method that uses the shadowing planes and the basis images in mixed reality (MR). The method is model-based and computable in real time by using a standard graphics processing unit (GPU) making it particularly effective when used with an outdoor MR application. To express the shadows of objects, we map the shadow images that are synthesized from the basis images rendered offline with the basis lights onto the shadowing planes, and we compute alpha blending with the objects. The shadowing planes are generated from convex hulls of objects. We obtain the radiance parameters of the real scene from an omnidirectional image and compute the linear sum of the radiance parameters and the basis images on the GPU. We applied this shadowing method to both an indoor and outdoor MR application and found it to be effective.

View full abstract

We constructed a computer-generated hologram for a three-dimensional image reconstruction of a virtual object. We used an error diffusion method to improve the accuracy of the reconstructed image. Selecting a diffusion coefficient influences the image quality of the reconstructed image. Therefore, many researchers are investigating the optimization of coefficients. We estimated coefficients using statistical analysis. Using regression analysis, matching coefficients and object positions is possible. Calculation time and reconstructed image quality using the proposed method are better than SPA. This paper presents the proposed method and the reconstructed image.

View full abstract