Compared with the conventional endoscope attached CCD camera at the tip, fiberscope has some advantages, i.e., we can reduce the diameter of the tip for imaging drastically, and we can flexibly capture image by bending it along with the portion to be observed. Therefore, fiberscope is expected to be a new imaging device that increases the possibility supporting diagnosis and operation in medical front. In particular, if we want to capture image inside a very narrow tube like blood vessel, super-thin fiberscope is quite useful. However, the color information in the image captured by fiberscope depends on the number of fibers, since one single optical fiber can convey only one color information. Thus, the less number of fibers in fiberscope, the more color information is lost. Also, honey-comb structure noise caused by the clad which does not contribute to transmit any color signal seriously deteriorates the quality of image captured. In this work, we study on the method of the super-resolution to improve the resolution of images captured by super-thin fiberscope with the limited number of fibers about 5,000 fibers.

View full abstract

Different from endoscope such as gastroscope attached CCD camera at the tip, fiberscope can form an image outside body, which makes easy to thin the insertion part of it. Therefore, fiberscope is expected to be a new imaging device that increases the possibility supporting diagnosis and operation in medical front. Optical fiber used in fiberscope isolates light within each fiver by covering the material called “core” showing high refractive index with the material called “clad” showing low refractive index. Fiberscope is constructed by bundling a number of this optical fiber. However, it is well-known that the quality of images captured by fiberscope is seriously deteriorated due to honey-comb structure noise caused by the clad which does not contribute to transmit any signal captured. In this work, we study on denoising for images captured by a super-thin fiberscope consisted of the limited number of fibers about 5,000 fibers to improve the quality of images captured by it to the level necessary in diagnosis and operation.

View full abstract

The ratio of elderly people whose age is more than 65 occupied in the entire population of Japan reached to 25% in 2013, and the ratio increases year by year. Human bodily changes by aging such as yellowing of lens, pupil size reduction and so on, and from this reason elderly people becomes to perceive deteriorated color than young people. From this point of view, a color reproduction method which simulates image that elderly people perceives on PC, and/or a color correction method to make the image that elderly people perceives close to the image that young people does have been proposed. However, in the results of artificial color perception test using SPP3 through the conventional color reproduction method by elderly people, there are a few mismatches with the results of actual color perception test for elderly people. In this report, to improve reproducibility of color perception by elderly people and make the results closer to the clinical data, we propose a new simulation method model that newly employed tritanomaly in conventional color perception model and verify its effectiveness.

View full abstract

In this study, we studied cached video content management methods in video content services for multiple quality in order to adapt to environmental changes of content usage, e.g., an access variation. The conventional method used the number of content access in the reference period. The access prediction method using multiple reference periods was proposed. In addition, a transition rule for hierarchical video was defined. Next, as content usage models, delivered content model, access variable model and quality variable model are defined. By simulation experiments, we considered the efficiency of the proposed methods for the conventional method.

View full abstract

In this paper, we propose a new detection method for GGO regions from LIDC database based on image features.

View full abstract

In this paper, we present a method to perform digital communication of an interference fringe type computer generated hologram using IrSimple. IrSimple was defined according to the IrDA technical standard and was developed to send image data at high speeds using infrared light. We performed visible light digital communications using the uni-directional mode of IrSimple to transmit a 3D data bitmap file. Visible light digital communication was performed using a modified IrSimple system. The results showed that the modified visible light communications system are performed well.

View full abstract

In this paper, as an improved technique to produce more durable reconstructed images with smaller fluctuation level, we studied the constructing technique of soap bubble screen. As this result, more durable projected 3D images are observed with full parallax than by previous techniques.

View full abstract

The electromagnetic diffracted field in the pyramid horn antenna is calculated under the Fresnel approximation. As a result, the complex hologram is made by diffracted electromagnetic wave on the computer. We considered that this is one of the computer generated hologram. Moreover, the wave source field is numerically reconstructed as a result of Fourier transform and inverse Fourier transform in the horn antenna aperture.

View full abstract

In this paper, we have developed a new micro bubble screen, and studied a fundamental technique which enables us to apply this micro bubble as a spatial screen. As this result, we have suggested that a good improvement can be performed on the characteristics of the images and in the projecting system.

View full abstract

In the study of generating human behavior in a virtual space, we have developed a method to memorize branches and landmarks that are located around the branches as a cognitive map by using visual information. As the result of the simulation that generates walking behavior based on the created cognitive map, we have reproduced the same walking behavior as the original one, and confirmed that the memory is generated correctly.

View full abstract

This article is a consecutive study on automatic generation of portrait by using neural network. The previous study has the problem that the generated portrait was not very similar to the face photograph. This paper proposed an algorithm that only focuses on researching the outline of jaw instead of entire face. Besides, the method of eye shape generation has also been changed. From the results it is proved that the similarity between generated portrait and face photograph has been greatly improved .

View full abstract

The picture assessment generally requires the subjective assessment. However subjective assessment would be difficult including contradictory relations such as noise and resolution. The subjective assessment considered the relation was reported for the landscape pictures. However, one of the most common pictures of human faces have not been studied. We propose a novel assessment method in this presentation. According to our research it gives repeatable results and good performances.

View full abstract

When a large-scale disaster occurs, collecting and providing disaster relevant information on such as evacuation warnings is quite meaningful for preventing the damage from spreading. It is known that when the Great East Japan Earthquake occurred in 2011 in Japan, numbers of people utilized Twitter and it was reported that approximately 80% of Twitter users regarded Twitter as a utilitarian tool for gathering emergency information. However, it is difficult for the people who are not familiar with the use of Twitter to report disaster situations accurately. Moreover it is also difficult for them to gather information by utilizing Twitter. Then, in this study, we construct a Twitter client dedicated to reporting disaster damage situations with a function of attaching geotag (longitude and latitude information), hashtag for reporting disaster situations which indicates the municipality of the user’s current place, and address based on the position information of the user automatically. We also construct a system which displays disaster situation reporting tweets in the neighborhood of user’s current position by plotting on a map.

View full abstract

There are two types for AR (Augmented Reality); one is the location-based AR which requires location information, and the other is the vision-based AR which uses image recognition. Moreover, the vision-based AR can be classified into two types; the marker type whose recognition object is characteristic pattern often called the marker and the markerless type whose recognition object is actual objects. The vision-based AR is more popular than the location-based AR because the location-based one has the disadvantage in that the performance depends on the accuracy of sensors like GPS. In the case of the marker-based AR, the thing that exchange the marker is necessary when switching a display content sometimes causes the inconvenience of users’ operability. Therefore we are aiming to realize markerless AR whose recognition object is hand which can express many patterns easily. In this study, we focus on hand shape recognition, the element technology of the markerless AR, and propose a hand shape recognition method which is suitable for the markerless AR. We conducted an evaluation experiment to evaluate the proposed method. Experimental result shows that the proposed method has the effectiveness for the practical use.

View full abstract

Evacuation training simulator based on the walk-through in the virtual space is suitable for stacking up pseudo evacuation experiences. However, it usually takes time to master the operation of moving in the virtual space. This is considered to be a fatal problem at the event having the limit of time such as a disaster reduction workshop for general citizen. From the different view paint, it can be considered that the operation of moving becomes simple if the moving path is limited to the rail which shows no room for user’s selection. Therefore, an automatic route generation method for setting rail in the virtual space is proposed in this paper.

View full abstract

The number of the people who died because of the cerebrovascular disease is 118,347, which is 9.3% of the total number of the people who died in 2014. Equipments utilized for medical treatment and inspection of the cerebrovascular disease include MRA (magnetic resonance angiography), CTA(Computed Tomography Angiography) and DSA(digital subtraction angiography) ,which can generate images of blood vessel. To precaution and detect early-stage disease, CTA and DSA cause the problem of radiation exposure and radiocontrast-induced renal failure. Thus, these method cannot be applied for these purpose. In contrast, MRA does not cause these kinds of problems and have no impact on human body. Therefore, MRA can play an important role for precaution and detection of early-stage disease. However, the strength of the signal of MRA highly could depend on the speed of blood. We make a research for the purpose that precaution of the cerebrovascular disease by utilizing MRA to analyze the 3D structure of blood vessel of the Internal Carotid Artery. For this purpose, the reconstruction of 3D models of blood vessels based on MRA data is necessary. The reconstruction of 3D blood vessel based on the threshold processing is difficult, because it highly depends on the speed of blood as previously mentioned. This paper reports the result of comparison between MRA data and simulation of Internal Carotid Artery to make clear how to solve the problem of the dependency of the blood speed.

View full abstract

Analysis of cellular behavior is signif icant for studying cell cycle and detecting anti cancer drugs.Isolating individual cells in confocal microscopic images of non stained live cell cultures is very difficult process. Because these images do not have adequate textural variations and the cell s are often overlap in consecutive focal planes. Manual cell segmentation requires massive labor and it is a time consuming process. This paper describes a automated cell segmentation method for localizing the individual cells of Chinese hamster ovary cell culture. K means clustering method is used to extract the cellular regions from the background. Subsequently, level set method is used to determine the boundaries of individual cells. The result indicates the significance of the proposed method as it appr oximates the boundaries of cells, which are placed very close r or overlapped in each other.

View full abstract

In this pape r, we proposed a segmentation method to the fetal in ultrasound im ages using SuperVoxel and GraphCut. We use a SuperVoxel algorithm, simple linear iterative clustering (SLIC), to cluster pixels which have near values and similar 3D coordinates into a SuperVoxel. We generate a graph using the SuperVoxels clustered by SLIC and cut the graph to get some parameters which are used to check whether the baby is healthy. We test our segmentation method in the ultrasound images which are taken from a phantom The result of the segmentation is quite good and the accuracy rate is re ally high.

View full abstract

Image segmentation is considered as significant process in image processing. Since output of image segmentation can be applied in various areas. Several well known conventional approaches of image segmentation such as Otsu and Kapur tend to use lot of computation time which grows exponentially as the number of cluster increases. Thus recent study of image segmentation uses swarm based algorithm. The swarm based algorithms stand out for their ability in generating high quality solutions for solving various problems. Among swarm based algorithms which are commonly used in image segmentation, Artificial Bee Colony (ABC) is considered as one of the best algorithm due to its convergence speed and segmentation result. Most of prior studies define the number of image region in initialization phase. However, identifying the correct number of image region is not a trivial problem. Since the high diversity of image region in real image, thus the correct number is not known a priori in real image. This research propos es novel method for image segmentation using multi level ABC. The first level ABC segmentation aims to define temporary set of threshold for image segmentation. Whereas the second level ABC segmentation will refine thresholds of the previous step. Between those phases, the number of threshold which divide image into region will be optimized. The proposed method shows to improve image segmentation quality within faster computation time.

View full abstract

In this paper, the authors proposed a method to estimate the motion of any point on the surface of the tree from the video sequence by using few feature points. More specifically, the center lines is obtained by trunk extraction from the video sequence, and the intersection candidates among the branches are detected. By matching the intersection candidates between consecutive frames, the intersections are screened from the detected candidates. The branch between each two consecutive intersections is seemed as one segment and any point on the tree surface is clustered into the nearest segment. Based on the motion information of the segments, the motion of the points could be estimated. The experiments prove that the proposed method has the good performance for CG animation of tree.

View full abstract

This paper proposed a method that can determine whether a target object can be cleaned-up by Clean-up Robot. The target object can be detected from the 3D Point Cloud Map which is generated by utilizing Simultaneous Localization and Mapping (SLAM). This method uses SLAM to generate two sets of 3D Point Cloud Map, the first 3D Point Cloud Map is a scenario of an indoor environment, the second one is the same scenario adding target object that need to be cleaned-up. The detection algorithm of target object used the differential point-cloud data of these 2 maps. Then, 3 kinds of filters (1. Statistical filter 2. Coordinate filter 3.Radius filter) were used to remove the noise. After the target object was detected, Clean-up Robot is used to push the object. By estimating the distance of movement of the target object with 3 ways (a. Robot’s Odometry sensor, b. Visual Odometry, c. Differential data of 2 maps), we can finally determine whether the object can be cleaned-up. Experimental results carried out with actual images show that the method of this paper can be applied on Clean-up Robot in the future.

View full abstract

In this paper, we proposed a novel method to estimate the focal length when camera motion satisfies Extended Standard Vergence (ESV) constraint. A simple method for achieving manually ESV constraint in real world was also provided. Both synthetic and real data experiments showed that this method could provide more accurate estimation.

View full abstract

[in Japanese]

View full abstract

This paper proposed a method that can early recognize suture surgery actions before they finish. The proposed method also can calculate the percentage of action completion. The experimental results proved that, the proposed method could early recognize surgeon's actions during suture surgery effectively.

View full abstract

Super Resolution Image Reconstruction (SRR) is one of the most popular Super Resolution (SR) algorithms and many studies are reported. Although SRR uses complex and difficult mathematical equations, the principal is the same as the method of least squares. In this presentation the truth of SRR is reported and the limitations of SRR is discussed.

View full abstract

This study proposes a method for measuring the spatial position of objects using wide-area active cameras in a surveillance system. In order to measure the spatial position, it is necessary to estimate the external parameters of cameras. In this study, only the information on the optical axis is used to estimate the external parameters, so that the accuracy is hard to be influenced by the lens distortion, and the position of objects can be measured with high accuracy.

View full abstract

Big data analysis for creating the new value by analyzing various data generated by people and things has become a hot topic. In this talk, I introduce typical machine learning techniques for big data analysis and show what kinds of big data analyses have been recently done in several important applications.

View full abstract

In this paper, present conditions and prospects on information monitoring of buildings are described. At first, present conditions on information monitoring of buildings are described including present ICT environments and research results by the authors are presented. Finally, prospects on information monitoring of buildings are also described.

View full abstract

We investigated the spread of the auditory presentation code (SP code) in local government.We announce the purposes, the methods and the results of these investigations and also the problems to be solved.

View full abstract

I will share the current situation of VR devices which will be launched in game industry from my personal coverage in GDC2015 and E3 2015.

View full abstract

As other media contents, manga has been consumed through digital media, and it becomes usual that they are produced and distributed in the digital environment. Therefore, it is a crucial problem to manage various information resources of manga which are increasing rapidly. Manga is a content to express complicated information by the mixture of pictures and texts. It is difficult to access and use this information from a simple graphic data. This obstructs the efficient use of manga in the digital environment.We are making some researches for the efficient management of manga information resources by semantic web technologies. this speech introduces them, and additionally, refers the expectations of the growth of research about manga in the research area of media computing.

View full abstract

Recording, updating and managing the historical information of buildings are difficult as compared to other industrial products. This paper reviews and discusses current situation and future perspectives of management of building historical records, with special focus on the digital construction photography.

View full abstract

In the field of fish population dynamics at fisheries science, the information technology has been how to use, and also will be discussed about future possibilities.

View full abstract

It is expected that a single life old man (target person) will be the people of 6 million soon, and a problem becomes serious with this increase in the near future. Care demand increases, and the lack of the helper is caused by acceleration of low birthrate and aging. There is this serious problem that should be settled in the Japanese society of the near future. This study is that I think about these measures, and an old man can continue living in peace at the place that lived so long in the old age. Like a human being, it is that it is made a help without an old man standing alone. It is to have you send relief and life of the security. It is planned to switch over from a collective nursing facility to the care of individual houses as a policy of the countries. About the measures, I think that it will become useful. I aim at becoming the generic technology to support an old man by low birthrate and aging by ICT and a fusion technology of care robot (RT) of RT. I fill the gap of a family and the helper visit and measure the health condition of the daily old man with a sensor and a device with a thing aiming at “the care without interruption” for 24 hours by, IaaS. I want to in this way try care to foresee a disease by accumulation of Vital Signs，when care RT gathered it beforehand more to support before an accident occurred, and to decrease an accident. Of course I press him and her for correspondence with a sound and an image by measures immediately and deal promptly again when the temperature rise of the old man was examined. Environment without the everyday accident must be provided to the old man. For the solution, it is to cope with an environmental change of the care quickly. Through development of these care RT, it will become the advanced home health care. Therefore, with the development of application and the OS, the way of the urgent notice and report may be necessary, too. In addition, I unify healthy information by computing, and big data should be able to utilize the care robot by becoming it. In a care service, reliable safe gentle old age society is seen.

View full abstract

The scope and the background of the session tile “Local Community Activation through Intercultural Communication” have been described based on the commoditization of information technology. Necessity of dialogue between engineering professionals and human social professionals is proposed and the possibility of the collaboration between IIEEJ and SIETAR Japan has been discussed.

View full abstract

Nowadays, women are expected to play the active role in Japan. However women’s jobless rate is still higher than men and the number of women who work as management level is limited. In order to break the situation, starting new business is considered to be one of ideal solutions for women who could not find the job at corporations. In this study, matrix resumes of Yae Niijima and Umeko Tsuda are reviewed in terms of entrepreneurship. As a result, there are commonalities between them in career anchors, enthusiasm for women education and strength to overcome difficulties.

View full abstract