In this study, we analyzed and quantified the three - dimension a l shape of replica skin to evaluate skin texture, and then investigated how to apply these techniques to quantify skin textu re based on two - dimensional microscope images of the skin.

抄録全体を表示

Combining CNNs for object detection and semantic segmentation is effective for a scale-aware bird detection. However, a CNN for semantic segmentation requires many images with pixel-level annotation during training. We propose to train the CNN from images with bounding boxes and compare our method to the strongly-supervised counterpart.

抄録全体を表示

In recent years, Deep Learning has been focused on not only in image recognition but also in various fields. In this paper, we applied DCGAN to CNN’s data augmentation. By the experi ment, w e confirmed the effectivity of this data augmentation and the recognition success rate reaches to 7 0 .6% .

抄録全体を表示

Conventional designing methods for multispectral filter array require training images covering a wavelength range of reconstruction target. We propose a new designing method which determines the measurement wavelength and pattern of filter array without using images covering the target wavelength range.

抄録全体を表示

In this presentation, a stereo tracking system composed of two 1ms Auto Pan/Tilt systems is reported. Real-time 3D position estimation of the target and robust tracking function were demonstrated to show the validity for sports training and sports broadcast.

抄録全体を表示

Visual tracking of a high-speed ying object is required to study the aerodynamics around the object. We propose a new system to track a high-speed ying object so as to achieve higher resolution imaging and substantiate the validity through a tracking experiment.

抄録全体を表示

We report speeding-up of calculation of the reconstruction processing in digital holography in order to realize realtime multispectral digital-holographic high-speed imaging flow cytometry. We adopted digital holography which applies spatial frequency-division multiplexing method to recording of red-, green-, and blue-color holograms. Because the reconstruction processing takes 1.6 ms by using GPU computing, we estimated from the calculation results that the imaging- flow-cytometry system could realize 2,500 cells/s as measurement throughput.

抄録全体を表示

We focus on a practical problem when a programmable aperture camera is used to capture 4-D light eld data. Different from an idealized theoretical model, the aperture plane is not uniform in light transmittance. To tackle this problem, we propose to measure the relative transmittances through a calibration process.

抄録全体を表示

Toward the migration to a next-generation system, we are considering the proposed specication for terrestrial broadcasting. The proposed specication inherits the advantages of the current system and various parameters are modied. It also incorporates the latest technologies, such as LDPC code and non-uniform constellations, and improves exibility and spectrum uti- lization efficiency. This report presents an overview of the proposed specication and the results of laboratory tests using prototype equipment.

抄録全体を表示

We are developing a next-generation terrestrial broadcasting transmission system. And we consider a study of partial reception for improving the reception performance and increasing the transmission capacity. This paper proposed a method to boost only partial reception.

抄録全体を表示

We have been developing a transmission system for next generation terrestrial broadcasting that enables 8K Super Hi-Vision. Regarding the carrier modulation scheme, we have been studying non-uniform constellations (NUCs) which improve the transmission performance. In this study, we verified that the improvement in the required CNR corresponds to the improvement in the CNR theoretically calculated from the BICM (Bit-Interleaved Coded Modulation) capacity and that the improvement becomes greater as the bit/symbol increases in the carrier modulation.

抄録全体を表示

Propagation over terrestrial broadcasting tends to receive multipath from buildings or mountains. Therefore, ISDB-T adopts OFDM with forward error correction and guard interval. We are conducting research on digital terrestrial broadcasting to enable large-volume content services such as Super High-Vision, and considering to adopt LDPC code to the forward error correction. Evaluating the performance of forward error correction not only in AWGN environment but in multipath environment is also important for terrestrial broadcasting. This report presents the evaluation results of transmission performance of LDPC coded OFDM under multipath ensembles, and compare with transmission performance of convolutional coded OFDM. The results show that LDPC code can improve the transmission performance in a severe multipath environment.

抄録全体を表示

NHK is conducting research on the next generation terrestrial broadcasting to provide large-volume content such as Super Hi Vision. The proposed specification for the next generation terrestrial broadcasting has designed based on segmented OFDM to provide mobile reception service the same as One-seg service. This paper describes a computer simulation of mobile reception performance to compare the proposed specification and ISDB-T.

抄録全体を表示

We made a spatial optical modulator using liquid crystal device and retro - reflective sheet in down - link direction for request signal. It can eliminate light sources in down - link and save power consumption . Extinction radio of more than 15dB was achieved over incident angle of ± 25 degree .

抄録全体を表示

We propose an underwater optical transmission system that sends video from a submarine to ship via an optical repeater. We suggest a new method for detecting misalignment of optical beam easily by covering the surface of the repeater with a retro-reflective sheet. We succeeded in irradiating the repeater with tracking beam autonomously by conducting pan-tilt operation. We found that the allowable angle setting error among four optical transmitters was around half width at half maximum of an LED module.

抄録全体を表示

We developed super high precision type FM modulator designed by all digital processing from Audio input to modulator output, which enables SFN(Single Frequency Network) of FM Radio broadcasting

抄録全体を表示

Recently, elderly persons with dementia h ave been increasing, and it would be one of the severe social problems in Japan. In the field of geriatrics, it is usually said that detecting dementia in the early stage is very important. The authors have been discussing method s to evaluate a patient’s dementia using clock drawings and simple graphics. This paper discusses a method to extract sides from given graphics for features extraction.

抄録全体を表示

In this report , we investigate the infl uence of network delay on human perception of weight with haptic sensation in a virtual environment. In our experiment , two user s have both end s of a stick with a weight and lift the weight in a 3D virtual spac e while trying to keep the position of the weight in the center of the stick by haptic interface device s . If the stick slopes, the weight slips along the stick and the weights of both ends of the stick change . W e also make the experiment in the cas e where the users cannot see the virtual space but perceive only the weight

抄録全体を表示

The representative depth distance was examined for the integral photography images, which were generated from the multi - view images in the case of setting the fixation point.

抄録全体を表示

Previous studies have suggested that image resolution and contents have little effect on the preferred viewing distance (PVD) of traditional TV systems. However, in future TV systems such as ultra-high-definition (8k) TVs, which are expected to have a wide field of view and high spatial resolution, TV program production methods may change depending on the system parameters, and viewer behaviors including PVD may change accordingly. Here we show that PVD is shorter in the 8k TV than the high-definition TV.

抄録全体を表示

We propose a method about space information analysis and visual effects for sports scene using multi-viewpoint robotic cameras. The method is composed of three elemental techniques (multi-viewpoint robotic cameras, object tracking and camera calibration) and possible to express subject motion in 4D space (3D space and time) clearly. In an experiment, we generated new visual effects for sports scene and confirmed the effectiveness of the method .

抄録全体を表示

We developed a prototype 3D ball tracking system for sports graphics. It is consisted from 3 technology elements, camera calibration for multi-view cameras using angle sensors, 2D multiple ball tracking, and realtime triangulation. Each technology built up for realtime system, and connected to 3D graphics system in order to feed the data. We examined on small field and confirmed the effectiveness of the system.

抄録全体を表示

The video shooting techniques differ in competitive sports from different various characteristics. In this talk, movies and their shooting techniques required by national level athletes in Japan for sports individually will be presented. These techniques can be used for the skill improvement for athletes.

抄録全体を表示

We have developed an application to support the practice of basketball formations. It can analyze videos of the practice, detect all players’ positions, and automatically generate 2D animation of all players’ movements. Our application is useful to improve the efficiency of the practice of basketball formations.

抄録全体を表示

We study on 64APSK for improving transmission capacity by using 12GHz band broadcast satellite. In this report, we show the way of bit allocation by maximizing channel capacity to apply Set Partitioning method. After that, we indicate the distribution ratio of LDPC coding rate estimated from C/N vs BER simulation before the error correction.

抄録全体を表示

This paper describes construction of FPU receiving base stations with automated helicopter tracking system and remote control system. These Facilities give us television program contribution from the entire area of southern Wakayama.

抄録全体を表示

To develop a bidirectional mobile MIMO FPU system, we must grasp propagation parameters required for simulations of propagation paths with fading. We carried out propagation experiments in the 1.2-GHz/2.3-GHz bands with mobile FPU systems in urban/suburban areas and grasped the parameters.

抄録全体を表示

TDD ( Time Division Duplex) - base d bidirectional digital FPU(field pick - up unit) system is now being stud ied . We studied the frequency sensing schem e using SINR (Signal - to - Interference plus Noise power Ratio) an d Er ror Rate in data transmission duration . From the results of indoor experiment , it was confirmed that the proposed scheme was useful.

抄録全体を表示

In this paper, the head pose estimation method for football videos using HOG and color histogram features was studied. The method classified low resolution head image into 8-class head poses. It was evaluated with football video taken by a single fixed wide-field-of-view camera and showed that supports real time processing.

抄録全体を表示

A method of perspective distortion correction for photographs taken of Japanese characters on signboard is discussed. Two vanishing points of Japanese characters computed using Radon transform. A homography matrix is calculated using the vanishing points, and its perspective distortion is corrected by the homographic transform.

抄録全体を表示

Conversion for blending the color signal and enhancing the color vein image for visible light photographic vein image are discussed. Coefficients in our method are optimized using the simulated annealing.

抄録全体を表示

I t i s required to recognize little color difference which human being cannot distinguish by using a standard camera. The purpose of this study is to determine threshold between indistinguishable color difference and distinguishable one . As a result, it is clarified that the threshold is determined by optimizing coefficients of ⊿..00.

抄録全体を表示

A video streaming service is getting available at all times and all places. DASH is a common protocol for the video streaming and various technologies to control it are proposed. Most of them use the past throughput to when and what bitrate to select. However, the throughput varies dynamically especially on the move taking the train or the bus and predicting the throughput variance in advance is difficult. So relying on only the past throughput leads to improper control for streaming. We investigate the throughput variance caused by the moving state and propose a novel technique considering the moving context. Experimental results shows that our method reduces the video stalling time by 40%.

抄録全体を表示

S ocial graph analysis for nursery schools is discuss in this paper. The social graph is created by a glasses type video camera wearing nursery teacher s and a face authentication sys tem. Some kind of centralities on each child are computed from the social graph , and we can estimate some important children who are reader in the group.

抄録全体を表示

Automatic generation of TV-program like CG animation (CGA) has been studied. We have developed appropriate show formats applied to the dialogues extracted from the bulletin board. An evaluation test is performed by comparing the original text contents and the converted CGA to identify the advantage of our method.

抄録全体を表示

We propose a new communication tool by using TVML(TV program making language). There are some difficulties to have good communication only by text messages. Taking advantage of character animation created by TVML, we aim to enriching communication with facial expression and gesture. This paper describes our system and results of subjective evaluation experiments of the system.

抄録全体を表示

In current education, active learning is very important to grow students. I n this research , we propose a new learning assist system using marking character area with a finger on a tablet terminal , and the marked character area is extracted to a sub - image. The sub - image is recognized by OCR. The correct recognition rate of individual character is 89.6%, and the correct recognition rate of every character in one mark ing area is 59. 3%.

抄録全体を表示

To evaluate effects disparity and resolution on the high resolution stereoscopic images, the preference of images with different disparities and resolutions were evaluated by paired comparison method. As a result, effects disparity and resolution were same degree on some images in the short viewing distance which viewers were able to tell the difference of resolution, on the other hand, the effect of disparity was observed only in the long viewing distance which angular resolution of image was over 30 cycle par degree.

抄録全体を表示

We precisely measured the vergence movement of the human eye When viewing a binocular 3D video image to obtain a stereoscopic image. If the Relation of parallax stimulus and vergence movement becomes clear, the mechanism of the influence of 3D video image on the visual function will become more apparent.

抄録全体を表示

In medical visualization, methods to recognize complex inner structures of 3D objects are important. In this paper, we investigate spatial perception effects when combining opacity tuning, opacity-based shading, and binocular and motion parallaxes of stereoscopic vision. We found out that proper opacity tuning and opacity-based shading improves human's spatial perception.

抄録全体を表示