A video search method is one of essential technologies in video content distribution systems. This study specially focusses on video identification as video search methods. A video identification method using the histogram of quantized motion vectors had been studied previously. However, it has some problems for the search speed because of a huge amount of characteristics. Thus, for the purpose of the higher speed for that method, this paper proposes video identification methods using the histogram based on each norm and motion direction group calculated from coded motion vectors. Moreover, the high-speed scheme, which uses each histogram step-by-step, is proposed. First, we considered the accuracy of identification in method (b) by the simulation experiments, and it was defined as highest accuracy to be satisfied. We confirmed that method (c) and (d) were able to identify the video queries, which have a number of unique features. By the evaluation experiments for the high-speed, we could also understand that proposed method (e) was about maximum 2.27 times faster than method (b) in the same accuracy as it, and consider about the efficiency of the speed-up as video retrieval methods and their applications. As further study, we will consider the proposed method for large-scale video contents DB in detail.

View full abstract

Popularization of wide area communication network realized ubiquitous computing society with high quality contents. These information distributing system based on wide area network is suitable from viewpoints of information security with resource effectivity, and so on, and was achieved as efficient information sharing system. This paper proposes novel digital rights management system based on information distribution and rebirth mechanism with redundancy dispersion method for systematic information re-construction, and describes technical consideration of its characteristics. The proposed method can control disclosure level and its license by information restoration processes and partial factors, and the user discrimination ability of proposed method can improve with cross-division and its interpolation process.

View full abstract

Recently, they had grown in frequency of use of video contents over the network due to the popularization of a lot of mobile devices. Then, video contents distribution methods using caching had been studied to realize more highly efficient video contents delivery. However, in a video management scheme using hierarchical data, if a calculating method of priority order uses the frequency of use, there were some problems of decreasing the efficiency of video contents management (VCM) because of the condition of decoding scalable data in which high quality data (H) absolutely requires low quality data (L). Then, this paper proposed a VCM method (m2) in consideration of the condition of usage quality. It adds the frequency itself and the high frequency for the priority value of low quality data. A quality-independent VCM method (m1) as the previous method, and order-shift method (m3) as low-quality-weighted VCM method were used. Moreover, the improved methods (m1', m2', m3') were defined respectively to reduce transmission data of each quality. We considered the effectivity of the proposed method for the condition of usage quality. We also compared the effectiveness of each method for data size of L and H. By the experimental results, we could confirm that m2' was the most effective, and that it was also important to consider both the quality-dependency condition and the data size. Next, a second caching method was used and we considered the effect of each method in variety of quality use. It was indicated that proposed method was the most effective if the usage quality was changed.

View full abstract

With development of cyber physical systems, information hiding technology that can go through the real environment has been expected, while some methods for embedding supplementary information into character patterns for improving recognition accuracy have been studied. If we change the design of character patterns drastically, it will become unnatural and readability will be lost. Supplementary information should be embedded as a natural design for humans, and it should be accurately extracted by using a camera. In this paper, utilizing the fact that human eyes are less sensitive to the change in color than in the luminance, we propose a method for improving recognition accuracy by embedding supplementary information hierarchically into character patterns by changing color of the region to some extent where it will not be visible for human.

View full abstract

Unauthorized recording in a movie theater for illegal distribution of digitized images is a serious problem. So far, information hiding techniques, such as watermarking, fingerprinting and so on, have been applied to prevent illegal duplication and distribution of digital contents. These conventional techniques can be useful for copyright protection in a limited situation of digital storage or network, however, can not inhibit unauthorized re-digitizing and duplication in the real world such as a movie theater. In this paper, we propose an unacceptable camcorder detection method, which is based on a geometrical distortion of 3 feature points on a screen. A technique like our proposal will lead to prevention against unauthorized recording.

View full abstract

We propose a video haptization system. Several haptic devices have been proposed and are commercially available. However, they are still high price and are used for the researches mainly. In this circumstance, we had proposed the haptic device called Spidar-mouse that can easily construct the interactive system. In this paper, we developed the video haptization system with Spidar-mouse by using the movement information of the object in the video and evaluated our system. The evaluation results demonstrated that our system effectively enhanced the understanding and fun of the video.

View full abstract

Technical demands for extraction of significant components from spatial models are increasing as 3D sensor and its application technology had been developed and popularized. Conventional feature extraction for 2D image cannot apply to depth images analysis, and is not suitable for 3D construction analysis because it extracts gradation between pixel values. In this paper, we propose a novel method for extraction of characteristic elements from 3D point cloud with an inspection about inner product distribution of normal vector on points. In our numerical experiments, characteristics on each point were extracted by the proposal method with robustness to noises and structural holes.

View full abstract

Recently, 3D and free-view displaying technologies are getting attention in movie and electrical industries. However, these technologies require to obtain a large amount of images with different viewpoints. The images are too redundant, since the images are almost similar to the neighboring viewpoints, although the expensive equipment such as large camera array and huge storage are required to capture and store the images. In this paper, we propose to capture smaller number of viewpoints and lower resolution of images, then we interpolate the viewpoints or resolution by image processing technique for the effective acquisition of multiple view images. We evaluated the image quality in objective and subjective experiments and discuss about the trade-off between the quality and effectivity.

View full abstract

This paper describes novel area-based stereo-matching method which aims at reconstructing the shape of objects robustly, correctly, with high precision and with high density. Our goal is to reconstruct correctly the shape of the object by comprising also edges as part of the resulting surface. For this purpose, we need to overcome the problem of how to reconstruct and describe shapes with steep and sharp edges. Area-based matching methods set an image area as a template and search the corresponding match. As a direct consequence of this approach, it becomes impossible to correctly reconstruct the shape around steep edges. Moreover, in the same regions, discontinuities and discrepancies of the shape between the left and right stereo-images increase the difficulties for the matching process. In order to overcome these problems, we propose in this research the approach of reconstructing the shape of objects by embedding reliable edge line segments into the area-based matching process with parallax estimation. We propose a robust stereo-matching method which integrates edges and which is able to cope with differences in right and left image shape, brightness changes and occlusions. The method consists of the following three steps: (1) parallax estimation, (2) edge-matching, (3) edge-surface matching. The edge detection procedure was described with details in a previous work. In this paper, we describe and explain deeply the process of parallax estimation and the area-based surface-matching with integrated edges; the performance of the proposed method is also validated.

View full abstract

In recent years, with the popularization of vehicles, road traffic has tended to increase. Therefore, it is important to rapidly collect road traffic information. In this study, we propose a method of determining the position and movement of vehicles and a method of showing the overhead view using images taken at an intersection from the oblique view with only one video camera. In this method, we use the vehicle shape model and determine the position of the model by the simple computation using the vanishing points and the vanishing line obtaining from the geometrical projection between the image and the road plane. Then, we project the position information onto the digital map, and we show the overhead view of the vehicle position in each frame of the image.

View full abstract

Roughly speaking, color reproduction of paper-based prints depends on the reflection of paper and the selective absorption of transmitted light of ink. As it is quite difficult to separate the ink layer from the paper completely, the thickness and transmittance of the ink layer cannot be measured accurately, and there is no precise model to represent the transmittance of ink layer of the prints so far. In this study, a single color solid print is treated as a bilayer model, composed of paper and single color solid ink printed on a transparent film, by ignoring the ink penetration in paper. At first, the transmission haze measurement is introduced to quantify the scattering character of the ink layer and the paper. Secondly, the transmitted light of the top layer medium is decomposed into collimated light and scattered light. Lastly, by superpostioning the collimated light and scattered light to the transmittance of the paper, the transmittance model of the bilayer medium composed of paper and single color solid ink printed on a transparent film was established. By comparing the predicted data of the proposed model, and the hitherto predicted data using Kubelka model with the measured data on various print samples, it is proved that the proposed model is much more precise than Kubelka model for transmittance prediction.

View full abstract

This paper proposes laser image binarization for crack detection by fractal dimension. Laser image is one of the road surface image and can be obtained by observing scattering light of laser from road surface. It has the advantage of robustness of lighting condition over an image that is taken by optical camera. However, surface asperity of road manifests as texuture regions in these images. So, crack detection method needs to be deal with texture regions. Although many of conventional methods geared toward concrete construct, these methods do not work well to road surface image. Proposed method can evaluate a texture region and binarized laser image effectively through the use of blanket method locally and globally.

View full abstract

This paper proposes a method for estimating in real time the pedestrian number from the variable blocks and prediction residual signal generated by the H.264/AVC encoder. A previously reported method estimates the number of pedestrians by using statistical methods or clustering after generated in a separate process once features from video, it was unsuitable for real-time processing. The proposed method estimates the number of pedestrians using regression analysis predicted residual signal generated by H.264/AVC encoder, and correct the number of pedestrians using the variable blocks. It was confirmed that it is possible to estimate in real time by using the data set of PETS2009 used in human tracking and background subtraction normally, the proposed method is obtained and performance equivalent to that of the previously reported.

View full abstract