In this paper, for the purpose of speeding up content acquisition and a reduction in service costs, we focus on the content management method using cached data for the quality in video content distributed system. The system model, which uses the frequency of access and the quality information, is defined. The management method of cached contents is also described. We use two categories, such as, the utilization of single quality and multiple qualities. At first, in a theoretical and an experimental approach, it is important that cached data is not the part data of hierarchical structure but the data of the quality which they frequently use. Next, we considered that the maximum caching hit ratio in this scalable method is 7.46 times higher than the priority method using the low quality, and the maximum inhibition ratio of transmission data is 0.53 times in comparison with the simulcast type. In the case of changes of access distribution, this method is superior to the simulcast type. Therefore, it was shown that the management method based on the independent frequency of the quality was effective in two categories, and the method using hierarchical data corresponding to each quality improved the efficiency.

View full abstract

This paper proposes a classification method of road congestion status using satellite images for collecting traffic information without ground-based sensors. In our method, we first extract vehicle edges eliminating several noises using various image processing methods. Then, we compute the number and the density of vehicles in each road section, and we classify the road congestion status using the density. In our study, we propose a new method of using three-dimensional map information for processing the occlusion and the shadow in the road caused by high buildings. We show the evaluation results of our experiments using real images in urban areas, and we show the possibility of traffic information collection in urban areas using satellite images and a three-dimensional map.

View full abstract

Aerial images are available in recent days. They are useful to observe the aging of the city. However, it's difficult to extract the significant spatial information from them. Many features in the city make it difficult to analyze spatial information by visual observation. In this paper, we propose two methods. One is a method to detect roof domains from aerial images using Multi-Level Slice classification. The other is a method which detects city geographical feature change automatically. In this paper, we run an experiment to verify the accuracy of our method and show the results.

View full abstract

Pedestrian detection required for numerous practical applications is one of the most challenging problems. Recently, accurate schemes for pedestrian detection have been proposed and achieved acceptable detection accuracy for several practical applications. However, the increase of computational complexity becomes a significant problem, because most recent schemes adopt sophisticated feature descriptors and advanced machine learning algorithms. To solve this problem, computational complexity reduction schemes and parallel implementations using hardware and multi-core processors are proposed, but the processing speed still remains insufficient for real-time computation. Considering this background, we propose an acceleration scheme that can be combined with existing schemes, and show experimental results using CoHOG-based pedestrian detection. In the proposed scheme, the number of sampling is reduced by efficient sampling based on the probability distribution computed from the results of sliding window detection at reference images. Experimental results using INRIA data set show that the proposed scheme can compute about 2.5 times as fast as the original implementation without any degradation of detecting accuracy where false positive per image (FPPI) is adopted as a measure.

View full abstract

In this paper, a binarization method based on fractal dimension for character string extraction is proposed. In order to deal with a scene image that has more than one character strings which have different colors (gray scale values), we have to take multiple threshold values. The proposed method can obtain multiple threshold values which correspond to each character string by detecting stable intervals of fractal dimension FD. The binarized character regions can be white or black, so we add reversed version of each character string region, that is, we generate “negative-positive problem” and then, obtain the black character regions by solving nega-posi problem. We also introduce two steps of noise reduction based on the shape of bounding box and edge information. The experimental result on datasets of ICDAR2003 shows that the proposed method has higher performance than existing binarization methods.

View full abstract

Writing is one of a behavior that can be used for biometric person authentication. A popular modality associated with writing is online signature verification. Online signature verification makes use of a trajectory of pen movement as a feature of writing. Though the trajectory is useful to verify signature, it occupies only a small part of writing features. In this paper, we propose a person authentication method using features extracted from pen holding style, which represents another aspect of writing. The features of pen holding style extracted in this paper are derived from both the hand's behavior and the hand's nature. Pen holding style is captured by a camera, and those features are extracted from the captured image. Dissimilarity scores are calculated from the features, and then combined in score level. In order to evaluate accuracies of proposed pen holding style verification, data from 30 persons are collected. The experimental results show that the proposed algorithm achieves the equal error rate (EER) of 2.7% against others' data, and EER of 4.1% against impersonation data.

View full abstract

When a user operates a robot, working in outer space or in nuclear reactors, the robot's view is sent to the user. To get the view needed to operate a robot, the work load will increase. To decrease the work load, we propose a 3D vision environment which the two pan-tilt cameras are controlled based on the eyes movement. The view recorded by two pan-tilt cameras is sent to the user and is used as the input of the 3D display. We introduce particles of eyes' size and likelihood using background region color space and attention region color space based on hue, saturation and intensity. The system estimates the eye movement of user and two pan-tilt cameras is controlled by using the user's eye movement. We show the effectiveness of the generated 3D environment for immersive and reality sense in comparison with a conventional remote system.

View full abstract

In order to project a large holographic 3D image in the space, a spatial screen of the 3D images should be required. For this, we have studied some projecting method with the use of mist screen to present holographic 3D images of moving pictures, and we have used water particles to reconstruct them. However, to make use of it as a stable screen, there appeared to be several problems to be resolved, for instance, the structural form of the screen is variable if the flow of the mist is disturbed, the projected image happens to disappear on account of unstable mist flow, and so on. In this study, to improve such undesirable situations, we have developed a new water flow(waterfall) screen made of liquid form of the water particles, and made efforts to construct more stable screen for the projection of 3D holographic images.

View full abstract

The content-based image retrieval system is a technique for retrieving images on the basis of automatically extracted features such as color, texture and shape. In this paper, we propose a similar image retrieval method by the wavelet packet transform for feature extraction of an image. The wavelet packet transform enables good approximation of an image because it allows better frequency resolution and much higher precision and flexibility in the selection of the bases. A query and all database images are transformed into the wavelet packet bases by the wavelet packet transform, and the optimal bases are chosen from the full-quad wavelet packet tree by the best basis algorithm. The structure of the optimal wavelet packet tree for similar images tends to be similar to each other. Our retrieving algorithm makes use of the structure of this wavelet packet tree and Euclid distance. The experiments were performed on the very large-image database containing 60,000 images from Corel Gallery. We evaluate the ability of our proposed retrieval method using a Recall-Precision graph. From this result, it is clarified that our proposed method provides greater accuracy, average Recall 25.96% and average Precision 60.0%. We also indicated that it is possible to easily obtain similar images from the very large-image database.

View full abstract

This paper proposes a UIMS (User Interface Management System) dedicated to home appliances, which is based on a script engine and can be executed on an embedded system with limited processing capability. The evaluation of proposed UIMS is performed from the view point of (1) the description and customization capability, (2) the program size and (3) the operational feelings. As a result, (1) by describing script les from about 1KB to 4KB UIs can be customized, (2) the UIMS size is 55.7KB and (3) UIs can be played over 30fps. The UIMS which can be executed on a low-spec hardware as home appliances is realized.

View full abstract

An investigation into encouraging higher user self-disclosure during text-based Computer Mediated Communication (CMC) system known as the Prompter Communication system (PCS), proposed by Yonemura et al. in previous studies, in which 1 to 3 black circular prompter images (“●”) are embedded onto the background of a text-field canvas as stimuli and as an embryonic basis for conversation, have been investigated and evaluated. The effects on supporting the projection or reflection of generated mental representations onto messages with the “●” prompter image as part of the representations within the message contents have also been examined. During the investigation, the concepts and features of the Geneplore cognitive creative thinking process model, together with self-projections, were integrated into the proposed system. Based on these concepts and features, the PCS aims to encourage users to express their mental representations by projecting or reflecting them onto the messages in combination with the provided “●” prompter image as part of the message contents. Communication experiments using the prototype PCS system have been conducted, where a Normal Communication System, NCS which uses only a plain text-field canvas without the “●” prompter image, is compared with the PCS. The experimental results indicate that communication conducted using the PCS results in an increased number of messages with the “●” prompter image(s) involved as part of the message contents than as background images. Further analysis into the self-disclosure level of the message contents and the relationships between the degree of self-disclosure and the representations of the “●” prompter image(s) within the messages have indicated that the degrees of self-disclosure increases when the “●” prompter image(s) are used to represent the participant him/herself, or the participant's thoughts and emotions, and decreases when the “●” prompter image(s) are used to represent animals or objects, or when the “●” prompter image(s) presented within the messages have been ignored. Investigations into the differences in self-disclosure level between message contents from both the PCS and the NCS have shown that using the PCS results in a generally higher degree in the average frequency of the appearance of high self-disclosure information than when using the NCS. During the conversation, these messages contain information with a high level of self-disclosure, which, consequently, may further lead to the development of intimate and deep interpersonal relationships.

View full abstract