By expressing the region in the wavelet sampling model, it is possible to reconstitute the sound-velocity distribution from the small number of propagation time. The small-scale experimental system for the demonstration was constructed. The effectiveness was confirmed by simulation and experiment. In addition, it was possible to obtain reconstitution image of which the approach accuracy is good by applying the triangular sampling to the sampling technique.

View full abstract

The tracking program was modified to track amoeba at cell division. The direction of each amoeba was determined by the vector from the common dark portion to each of the bright portions in the difference of the consecutive two frames.

View full abstract

Active Contour Model, Snakes, is one of the most popular boundary detection methods. Snakes has a problem that it is difficult to determine an initial contour and parameters suitable for the target shape. In addition, this method cannot detect the target boundary precisely when the target image does not have clear edges or uniform feature in the target domain and the background domain. We have proposed Snakes with decentralized cooperative processing. In this method, multiple Snakes do the detection procedure in a parallel and cooperative manner. As a result, the possibility of obtaining the target contour is increased by exchanging information between the multiple Snakes. In this paper, we present effectiveness of this approach through the experiments with simulation images and real images.

View full abstract

This paper proposes a face detection method using color and difference components in image processing. Extracting the regions of skin color and hair color, the face is determined from the region extracted as a moving object by using interframe differences. The movement of camera itself is considered in the detection of interframe differences. We constructed a real-time tracking system of moving person using this method, and it is confirmed that the accuracy of more than 90% is obtained from an evaluation experiment.

View full abstract

This paper proposes a discrimination method between trees based on the shape of leaves. Features are extracted by the Weighted Direction Index Histogram method from a binarized and normalized leaf pattern. Autoassociative neural networks are used in the distinction process. It is confirmed that the discrimination accuracy is improved by using several leaves of the same tree.

View full abstract

This paper proposes a pedestrian detection method using interframe difference and background difference in image processing. Pedestrians are detected by using the logical product of these differences, and the shadow area is removed from the color difference. We construct a real-time pedestrian counting system by determining each person from the contour shape. It is confirmed that the counting accuracy of 90.5% is performed in the evaluation experiment.

View full abstract

In this work, we propose an automatic ROI extraction method with Bayesian networks .We apply a the Bayesian network to semantic transcoding by taking into account ROI in spatial domain. In our experiments, we use football sequences because of easyness of modeling probabilistic structure in scenes of video sequences. We evaluate the performance of our ROI extraction method by setting faces of football players as ROI. Several experiments showed that the proposed method can achieve high recall and precision rate and can extract football players faces only excluded other faces automatically.

View full abstract

Recently, we developed high effective and high color rendering properties white LEDs lighting using near UV LEDs. In this research, we are described the making white LEDs lighting for medical fields, the evaluation of the lighting characteristic that uses the imaging camera system and the characteristic of the photographic image. And it's mounted that white LED on the top of electronic endoscope. Thus we introduce animal test for the over and under digestive tract of the beagle dog.

View full abstract

In this paper the algorithm which reconstructed the image from the small number of projection data is proposed. The region is expressed using wavelet sampling model and triangular lattice sampling. It became possible that the at high speed high-quality reconstruction image by formulating the relationship between sampling points and projection data, by using the Singular-Value Decomposition.

View full abstract

Stable feature extraction is needed to recognize nonrigid objects. Color Constant Color Indexing(CCCI) is a feature robust against deformation and illumination change, but its stability fails if the domain of the object is not segmented suitably. This is because the domain of the background is blended. In general, object segmentation is difficult if the background is complex. The method proposed herein uses a collection of local images and recognizes the object based on the distribution of the collection in the CCCI feature space. High performance recognition is achieved without object segmentation because we use the statistical characteristic that the feature vectors extracted from the background domain lie near the center of the distribution in the subspace of the object.

View full abstract

We propose a new watermarking method to embed a watermark using wavelet coefficients in several frequency bands and the sum of absolute value of two coefficients. In this paper, we evaluated PSNR of a watermarked image and residual ratio after the image was compressed by JPEG. We also compared them with the conventional method that utilized the vector conversion of the coefficients in the lowest frequency band of wavelet transform.

View full abstract

Along with the advance of the digital technology and the spread of communication network, the copyright protection of multimedia contents becomes more important. One countermeasure of this problem is digital watermarking. In this research, we found that our method which utilizes wavelet transform has a security defect, and we propose a new embedding method which improves the defect of our previous method.

View full abstract

We examined the lossless coding method of digital museum images. In the preceding studies, we studied the coding method using the rotation compensation with estimated shape information of the object. In this study, we propose the coding method using the rotation compensation based on perspective projection. In this paper, we examined the coding method combined the rotation compensation and the motion compensation.

View full abstract

In this paper, we propose a lecture video retrieval system using text information generated by speech audio signal. In this system, lecture video, audio signal, and scripts are distributed to users. The scripts are also utilized to find video data which users want to see. In addition, we studied on function, which enables students to join the improvement of the scripts.

View full abstract

This paper describes a robust technique for detecting color information by a camera-computer system. Visual sensing techniques are indispensable for realizing an intelligent robot having artificial eyes and a brain. When such a robot tries to understand its environment, it searches for shape, color, or texture information. Among them, many studies have been done to date for the purpose of detecting entities or identifying an objects using color information. However, color information is not a very stable feature to be detected, since it is normally influenced by illumination conditions caused by weather, time in a day, lights, shade, etc. In this paper, a technique is proposed for detecting color by the employment of two-dimensional color histogram. The technique is implemented on a pan-tilt camera and a computer system and it tracks a specified color object in a robust manner.

View full abstract

In medical rehabilitation training, it is requested to establish the method of evaluating training effect quantitatively. In this paper, a method of evaluating walking motion is proposed by asymmetry analysis of the motion. The proposed method is applied to the evaluation of walking motions of the patients suffering from coxarthrosis receiving rehabilitation training. Experimental results are shown.

View full abstract

The purpose of this research is to recognize 3D shape features of a human body automatically using a 3D laser-scanning machine. In order to recognize the 3D shape features, we selected the 23 feature points of a body and modeled its 3D features. The set of 23 feature points consists of the motion axis of a joint, the main point for the bone structure of a human body. For extracting feature points of object model, we made Spin-images neighbor for each feature points were extracted according to the feature points of the standard model. And the feature points were extracted by applying Spin-image matching. The extracted feature points can be applied as body measurement, the 3D virtual fitting system for apparel etc.

View full abstract

This paper presents a feature extraction method for the object recognition system that detects 3D position of object in real time and adequate accuracy from image sequences captured by camera. In our system, the object shape information is obtained by RFID-tags on the object through wireless communication. The presented method extracts feature points from the image sequences and compares them with the shape information to recognize the object position. The performance of the tracking method tested on a real dynamic scene was experimentally analyzed.

View full abstract

Recently, the trial to grasp traffic conditions automatically by using the template matching and so on is being attention. However, that contains a problem that can't get the precision of high distinction with vehicles' various characteristics. In this research, a new automatic traffic monitoring system using fewer feature points is proposed Minutia matching in biometric technique effective in finger point distinction is applied to the system. As an experimental result, superiority of the proposed method is confirmed by distinction rate compared with conventional method.

View full abstract

To determine 3D point of an object from multiple view images, corresponding points of each image must be known. In this paper, we examined a matching method to find the corresponding points automatically on multiple view images. In this method, both template of each image are decided according to an affine transformation by referring three known corresponding points, and new corresponding point is found by a pattern matching on the epipolar line. As a result, it was confirmed that right corresponding points are found on some images of object having plane and curved surface.

View full abstract

In a meeting, there is a problem that a minutes writer does not participate easily in the meeting. Therefore, there is a research by which the minutes making is supported. However, it takes the time of a meeting preparation beforehand. We studied a meeting state estimation which uses head movement by using multimedia meeting system for making meeting archive indexes. In this paper, we discuss a head movement collection method for four participants' real space meeting.

View full abstract

With the development of the Internet and digital technology, various contents have been made and used by all the people. Especially using a cellular phone with a digital camera, photo-contents can be made easily anywhere and anytime. In this paper, we propose a method of metadata description for the above photo-contents to distribute on the Net freely.

View full abstract

It is becoming more difficult for users to get needed information sufficiently and efficiently by using a cellular phone. One problem is that the interface function on the cellular phone is limited due to the required small size and weight, whereas it has many functions for various applications. In addition, it becomes hard to operate the functions with increase of complication of cellular phone system. This paper focuses on the information obtained from image sequences captured by cellular phone camera to understand the user context and assist the users to search information effectively.

View full abstract

Requirements of effective video retrieval is growing because of increase of video content distribution services. Therefore, in this report, we propose a feature descriptor that describe an object contour and its transformation. This descriptor is based on CSS(Curvature Scale Space) which is the shape descriptor defined in MPEG-7.

View full abstract

We present BROAFERENCE, a study on next generation multimedia home platform.The idea of BROAFERENCE is to add a video chat service to a common TV. Users that are watching the same TV program distributed via IP broadcast can lookup and invite others belonging to the same program group to build or join an adhoc video conference like chat. People can enjoy the TV program together and exchange opinions and informations by using the BROAFERENCE system. We present our approach for the problems of lookup and presentation for a huge amount of frequently changingparticipants as it will happen if persons switch between different TV channels.

View full abstract

Digital content distribution, which has been envisioned as a main industry of next generation, remains flat nevertheless the environment for dissemination of picture distribution has now established to some extent, and the path for the future growth also remains unclear. One of the possible barriers for dissemination of content is that each provider adopts different value chain partly because of the delay in the standardization of DRM (Digital Rights Management) scheme. The establishment of common DRM scheme is essential for system interoperability and terminal compatibility. This paper deals with a DRM standard, ISO/IEC13818-11 MPEG-2 IPMP(Intellectual Property Management and Protection), and extracts issues of the standard from two perspectives - implementation of the standard specifications and applicability of the standard to use cases.

View full abstract

The AV data and filesystems for digital AV recorders depend on the targeting standard of broadcasting and that of recording medium. We have developed a recording area management system that can share unused areas for two or more filesystems without dividing the recording medium into the partitions. And the system can treat any AV data format with real-time processing. And we have also developed a AV recording system that can manage the stream of the digital broadcasting format (MPEG2-TS) and the stream of the DVD format (MPEG2-PS) on the same hard disk drive (HDD) with our recording area management system.

View full abstract

[in Japanese]

View full abstract

It is becoming easy to create high quality digital cinema contents using video editing tools and to distribute it through broadband network. However, in order to promote digital cinema distribution, we think it is important that the platform is equipped with the function to search, recommend and to evaluate digital cinema. In this paper, we propose metadata model for digital cinema distribution, and show the overview of metadata circulation platform for digital cinema.

View full abstract

This paper shows an experimental result of an educational contents delivery using Digital Signage system. RemoteCastBB is a Digital Signage system and it allows store and manage signage data, consists of educational contents. A central server of RemoteCastBB is allocated at Kanagawa Prefectural Education Center and more than ten client laptop computers as RemoteCastBB terminals are installed at several high schools. Three kinds of content packages, those are signage data, are used to evaluate the performance parameters such as stability, consistency, onnectivity... of the system.

View full abstract

A reversible encoding of the image is examined. The band is divided after DPCM is first encoded as the technique, the image is divided into the high frequency element and the low frequency element, it aims at a suitable encoding for of each, and it aims at the improvement of the doing compressibility.

View full abstract

In this paper, we study on the cost evaluation techniques for mode selection in the video coding. In the video coding standard such as MPEG, it is significant to select the optimal mode from some candidate-modes in order to improve the coding efficiency. The Rate-Distortion Optimization (RDO) techniques is often applied to select the optimal mode. Although it is possible to select the mode that maximizes the coding efficiency, it needs expensive operations to calculate rate and distortion. We propose the method that accelerates RDO by using the RD model and the Zero-Block Detection. We implemented the proposed method on H.264 reference software, and demonstrate the effectiveness of the proposed method by experiments.

View full abstract

A lossy coding of error-diffused halftone image by coding a corresponding multi-level image with the same bit-rate as that of JBIG2 coding method, and error diffusion is done after decoding is studied. Various multi-level image coding standards are examined and image evaluation is done through inversed halftoning. The simulation result shows that proposed method is worth consideration.

View full abstract

This report presents the technique for the information hiding to cover-data of ascii-art. In this research, we tried to increase the capacity of hiding data with parameters control, such as the degree of complexity and the gradation depth, and the relationship between the image factors and the embedding ability were shown to the experimental results.

View full abstract

It is effective in contents understanding, such as object tracking, to analyze the feature not only in a pixel domain but in a frequency domain. However, it is necessary to decode and to convert again in a frequency domain for analyzing the image coded by MPEG. This paper proposes the method of tracking an object by using MPEG coded data. The experiment shows that the proposed method can track corresponding to change of a camera or an object.

View full abstract

In this paper, we propose a rewritable watermarking scheme on MPEG coded data domain for content managements including content controlling, indexing and DRM. Data embedding is performed in a block by block basis, where the length of zero run and the value of dummy AC component of quantized DCT coefficients are used as a data carrier. In the detection process, we can reconstruct the MPEG coded data that is very close to the original one, which enables us not only to rewrite embedded data but also retain the original MPEG video quality. In the experiment, we show that up to a few kbits/frame data embedding without any large PSNR penalty and data recovery can be realized using typical MPEG-1 coded stream.

View full abstract

We propose an advanced system design of a video streaming manual system supporting business education according to the examined system requirements. In this system, important objects are automatically detected, tracked and identified through spatiotemporal areas of video scenes and are spotlighted using several video effects in software video player with watermark detection. Object ID and additional information such as importance level, category, related URL and keywords are embedded into MPEG coded domain using the proposed watermarking scheme.

View full abstract

An anonymous payment system in network-type access control is required to keep anonymity of digital information users. Existing electronic money algorithm is lacking some elements of money-number, verification of double spending in using phase and cashback. We propose an anonymous payment system added their elements. Then we show a model that keeping anonymity of digital information users is possible by appling a proposed system. Keeping anonymity of digital information users is defined that conspired bank and supplyer can't get personal data from payed electronic money,and can't recognize which contents users used. At last, we prove the security of a proposed system.

View full abstract

This report proposes a new structure of image recognition and communication system to reduce total bit rate and operations by using the JPEG 2000 compression technique.

View full abstract

Universities and other academic organizations are now deploying campus public LAN that enables us to access educational media in lecture rooms and other area. Since, in schools, there are many people that are not belonging to the school, we must provide secure network accessing environment using authentication technologies. In this paper, we implement secure campus public LAN through combining the password authentication and MAC address authentication per individual users.

View full abstract

This paper, to create an effective rainfall scene, proposes the use of an algorithm that uses Level of Detail (LOD) . This method can reduce the stress of designers, creators, etc., enhance the realistic effect, and significantly reduce the costs of creating rainfall scenes.

View full abstract

Particle is one of the effective techniques that represents vague objects. This paper describes a method of representing mirage effect from the source image. Mirage is a kind of visual change phenomenon by air layers of different temperatures. First, we locate many transparent spheres between the view point and a source image. Then, tiny fluctuation of horizontal and vertical direction perturbs the position of these spheres. Finally, ray-tracing with only refraction is applied to these spheres. Phenomena of the mirage, such as reversal, expansion and contraction, are possible to control by changing refractive index in our method. In addition, it is possible to generate animation by changing refractive index and position of each sphere gradually.

View full abstract

Web3D has been developed to realize the three- dimensions visualization on web. It can be used to show the 3D contents on web to examine the surface and 3D configuration of scientific models. It has been applied for the rectangular magnetron system, the magnetic field by magnets, the 3D model of molecules, and the numerical maps.

View full abstract

The digital camera and it is widespread recently. Therefore, demand that processes the photograph has increased. However, the technology and experience are necessary to generate the illustration image with a photo retouching software. Therefore, it is not work that all members can easily do. Then, the purpose of this research is to generate the illustration image from the photograph image easily. It draws in the illustration image by using Area division and Color exaggeration.

View full abstract

Video game machine was cheap but people drawing row-res video game graphics used them brain. I takes them a long time to write row-res video game graphics. Video game machine have been powerful. For the reason, video game graphics become 3D graphics, but row-res video game graphics is effective for row-res display. I propose drawing edge method for row-res figure.

View full abstract

Much attention has been directed to an system that allows users to interactively view a scene in a cellular phone. Assuming simple scene consisting foreground and background objects for real-time processing, we propose a method that can segment each object region needed for the view generation. In the proposed method, capturing two images with different foci on objects from the same view point, we generate an all-in focus image from them and then compare the difference in intensity between three images to extract each region in efficient and robust. Error is also reduced by a simple non-linear filtering. The simulation results showed that the proposed method works in adequate accuracy enough for error is not visible in the finally generated free viewpoint image.

View full abstract

In performing texture mapping, higher quality pictures can be generated by using not only one picture but also multiple pictures obtained from different positions. Such texture mapping method is called a view-dependent texture mapping. In the process, the following points should be considered: whether a fixation point on the shape of an object can be visible or not from each camera. whether the camera's axis is close to visual axis or not. So far, in the view-dependent texture mapping, this decision has been processed on every each vertex in response to the amount of the data to be proceeded, and color interpolation was performed in the inside pixels. From this reason, satisfactory quality has not obtained for large polygons. The recent GPU (graphics processing unit) can provide users with more flexible processing than the former GPU by permitting the users to program the inside processing. This function is referred to as pixel shading. In this report, by installing the view-dependent texture mapping which judges the above criteria on every each pixel with use of this function, we compared the proposed method with the former method processing every each vertex unit. As a result, it was observed that the quality was improved by the proposed method for the case that the number of the polygons of the objective shapes is small. As an application of the present method, the reduction in the amount of the data in providing the 3-D contents which uses the view-dependent texture mapping via remote transmission, etc. can be considered.

View full abstract

DomeScreen is superior to a usual flat screen in the liveliness and a feeling of the depth. The various problems which occurred with the possibility were examined because it considered that the ultimate high liveliness display of the VR expression could be done if this system could be made without Glasses.

View full abstract