Digital watermark is a technology that protects digital intellectual property from being illegally copied and distributed by embedding it with information such as copyrights. By presenting the copyrights, digital watermarks serve as available technology for protecting the rights governing use of intellectual property. Recently, distribution of images taken from CRTs or other kinds of screens has become a problem. This is because current methods have difficulties detecting digital watermarks in the re-shot images. We present a method that can detect digital watermarks in moving images taken with a consumer camera from a CRT monitor connected to an NTSC signal system. The method can embed information in both temporal and spatial domains. It is embedded for a relatively long time and in low spatial frequencies to avoid degradation of images. The method can also detect digital watermarks in re-shot images that degrade in high frequencies. In addition, it can detect digital watermarks in real time without using the original images. This paper describes the method, presents an experimental system based on it, and verifies its effectiveness with various types of moving images. We also examine various aspects of the robustness needed for watermarks embedded with this method.
The current video switching/composition scheme employed is fundamentally subject to a variable delay of 0-1 frame at each stage due to the use of a frame synchronizer. We can fix and minimize this variable delay by controlling the video sync phase of the distributed video sources at the start of the communication session if they are driven by either the same video sync frequency or independent but very accurate sync frequencies. Furthermore, delivering the necessary phase controlling signals between the terminal and the video switching or composing point can be relatively easy because of the widely available IP networks. Based on these ideas, we determined the performance of an experimental system we constructed that used video phase synchronization of distributed video sources. We describe the system's architecture, discuss the sync source dependency of the system, and report the delay observed in existing and proposed schemes of video composition using real time hardware experiments. The results obtained indicate that the proposed video sync controlling system is effective and has practical applications for fixing and minimizing the video switching/composition delay.
Although a free viewpoint video can be synthesized using reference images with their depth maps, the quality of the output video can be degraded because of lack of information in occluded regions within frames. We previously proposed a high quality free viewpoint video replenishment system using a dynamically updated background buffer. However, this method cannot interpolate regions where the depth differs greatly for each point on the background, since the background buffer is approximated to a plane. We describe improving the method using a “curved background buffer” with depth information to achieve higher quality output video.
The need for ways to obtain highly accurate 3D shape data from existing 2D contours is increasingly necessary in some applications, such as in disaster prevention, environmental analysis, and creating construction plans. In this paper, we propose an algorithm that reconstructs a curve mesh model of mountain with triangular and quadrilateral domains that is generated from discrete contours data by using constrained Delaunay triangulation method. To prove the effectiveness of our method, we interpolated the curve mesh into free-form surface model with G1 continuity and compared the accuracy and quantity of data with other methods.
Normal maps are often used in computer rendering graphics to achieve more realistic objects. Both the number of normal maps and their data size have been increasing in recent times. Therefore, normal map compression algorithms, such as 3Dc, have been developed. However, there is no measure to evaluate the quality of lossy compressed normal maps. Therefore, the quality of graphic objects rendered with lossy compressed normal maps was evaluated by using the peak signal-to-noise ratios (PSNR) between the original normal maps and the compressed ones or by using the PSNR between rendered images, which is computationally expensive. We developed a model to evaluate their quality and to analyze the characteristics between lossy compressed normal maps and 3D images rendered with them.
A novel electron emitter utilizing the characteristics of ferroelectric ceramic materials has been developed. The electron emitter has a simple structure formed by using a thick film process, a small emission angle dispersion and its own memory function. The electron-emitting portion of the electron emitter has a unique microstructure. This micro-structure produces a novel two-step electron emission process by combining field emission characteristics with the ferroelectric characteristics of ceramic materials. To evaluate the electron emitter, a 4.3-inch prototype display panel was built that displayed a moving colored image using 128 gray scales per color.
We developed a technique to detect duplicated scenes, such as commercial messages, TV program theme songs, and repeated scenes from MPEG streams. Conventional matching techniques require numerous calculations to estimate similarity between scenes. In our technique, similarity is estimated with compressed code sizes without decoding pictures. We divided scenes into shots by detecting cut points, which probably correspond to large segments of generated code in MPEG streams. Pictures in these shots are not decoded, but only shot lengths are used for scene matching. If a series of shot lengths for scene A is the same as that for scene B, we can infer that both scenes are identical. This technique, called shot length matching (SLM), requires no image processing and works very fast. We applied SLM to 80 min. MPEG streams stored on an hard disk drive to detect and delete duplicated scenes, and we obtained 99.5% precision and a processing time of 0.157 s.
We have developed some human sensing systems that can be used in real time environments, such as on railway station platforms and at railroad crossings, to increase safety in busy public places. We describe an application for counting the number of people remaining in divided areas in large spaces. For this purpose, we installed stereo cameras on ceilings that point downward towards the boundaries of specified areas. The system can track all the people passing in each camera's view and can count the number of people remaining in each area. The system can work robustly even under extreme lighting variations by using the stereo cameras. It was used at “EXPO 2005 Aichi, Japan,” and was active opening hours a day for six months. We have stored all the tracking and counting data to make a precise analysis of the human behavior.
We propose an adaptive QoS control scheme for avatars which output voice and video streams in distributed virtual environments. The scheme performs processing load control, traffic control, and media synchronization control on the basis of the processing load of a terminal, the network load, and the importance of avatars. We also illustrate the effectiveness of the scheme by experiment.