We present a novel method for 3D object pose detection. Proposed method matches 2D keypoints against 3D model points directly and increases accuracy of the 2D-3D correspondences by three-step culling techniques. Experimental results show that threshold processing using viewpoint information improves the accuracy by about 35 points.

抄録全体を表示

For camera pose estimation from multiple markers in augmented reality systems, a relationship between markers generally needs to be known beforehand. In this paper, we propose a camera pose estimation method for multiple makers without prior information about maker layout, employing homography matrix decomposition.

抄録全体を表示

In the vision based camera pose estimate methods, the robustness is affected with accuracy of finding correspondent between feature points. In this report, we propose a robust method of finding correspondence points between images by using homography.

抄録全体を表示

The meta data while shooting is indispensable to effective picture processing and CG composition. So, we devised the new sensor which could measure angle and the position of the camera. We got precision around 0.2 degrees using MEMS sensor, and within 10cm in a rate of 30Hz by PIV method.

抄録全体を表示

In recent years, haptic technology has been widely researched and used in various interactive systems. Among these interactive systems, we are concerned about image haptization systems. In this research we propose a depth image haptization system with local deformation. The proposed system consists of a pc and a haptic device. A triangle polygon mesh is extracted and built from the depth image for the haptic rendering. Contact force between the contents and the user's finger is calculated with a proxy graph algorithm. In order to improve user experience, a local deformation is performed by applying displacement to the neighbor vertices of the contact vertex.

抄録全体を表示

In this research we evaluate two methods for haptic motion rendering of a dynamic image sequence or in other words a video, which enables the viewers to feel the motion of objects in the video beyond passive watching and listening using the string based haptic device, SPIDAR.

抄録全体を表示

This paper proposes an image projection system using a projector and a sphere-shaped screen. Several infrared LEDs attached on the screen are captured by a head-mounted camera for estimating 3D position and orientation of the screen, and a CG image corresponding to the user's viewpoint is generated according to the estimation result. By showing the image via the projector, our system provides the user with a feeling like there is a virtual object in the sphere-shaped screen.

抄録全体を表示

Augmented reality (AR) using a projector has recently been attracted. However, most those researches do not apply quickly and extensively dynamic objects such as an actor/actress in a play and a ball in a sports game but stationary objects such as a wall, a desk and a floor. This paper proposes projection-based AR for dynamic objects as a view of an application of high-speed optical gaze controller, Saccade Mirror.

抄録全体を表示

In this paper, we propose a VR-Blowgun-System with improving stability which resolves the safety and lifts several restrictions of usage environments. User assessment suggested this system make more enjoyable by using Virtual Reality and the same breathing way with blowgun. In the future, this system will be applied to clinic.

抄録全体を表示

In this paper, we propose a dynamic camera calibration and object extraction method for sport videos captured with a moving pan-tilt-zoom camera. Our proposed method estimates camera parameters by identifying reliable corresponding feature points between video frames, and also extracts objects' textures by projecting the feature line between field plane and stadium stand with estimated camera parameters. Experimental results show the effectiveness of the proposed method.

抄録全体を表示

This paper proposes a free viewpoint video synthesis method in wide space such as soccer stadium. The proposed method improves accuracy based on mutual feedback between tracking and segmentation for multiple objects. Experimental results revealed that proposed method successfully improved the accuracy of the boundary between multiple objects.

抄録全体を表示

We propose an ultrahigh definition video coding method with bit-depth reduction before encoding procedure and bit-depth reconstruction after decoding procedure. The bit-depth reduction uses Lloyd-Max quantization with image noise reduction and gradation preservation. The bit-depth reconstruction uses side information which is determined by comparing a local-decoded bit-depth reconstructed image with the original image in encoder side.

抄録全体を表示

This paper describes an inter-channel rate control method for an H.264/AVC 4-channel video encoder. The proposed method makes it possible to encode each video to the same extent subjective image quality by adaptively changing the encoders' target bit rates according to each video characteristics. The method can be used in various applications, including broadcast contribution and distribution. At BD-bitrate it improves the encoding efficiency obtained with constant bit rate coding by about 9.2%.

抄録全体を表示

We previously proposed a block-based adaptive spatio-temporal prediction method for efficient video coding. In this paper, we extend the method for decorrelating color video signals in RGB 4:4:4 format. In order to exploit inter-color as well as intra-color correlations, our method can use different color signals simultaneously in the prediction process.

抄録全体を表示

In the lossless coding scheme which was developed by the authors for RGB 4:4:4 format video, structures of linear predictors used for the respective color signals were different from each other. Therefore, it was difficult to determine the appropriate number of the predictors as well as arrangement of reference pels for each color signal in advance. This paper describes a method to adapt these conditions through iterative procedures of predictor design carried out in each frame.

抄録全体を表示

We previously reported on the still image coding scheme based on recursive type intra prediction. In this scheme, predicted values in each block were always generated in horizontal raster scan order. In this paper, we extend the method to allow the vertical raster scan order as well. Moreover, appropriate arrangement of reference pels used in the prediction is explored under the condition of adaptively switching both the scan orders.

抄録全体を表示

In the previous work, we proposed an image approximation method based on a linear combination of deformable continuous functions and showed its potential as a new image coding tool. In this paper, we investigate the statistical properties of parameters which describe shapes of the functions optimized by our method. Based on the results, we also introduce an entropy coding technique into several kinds of the parameters aiming at better coding performance.

抄録全体を表示

In HEVC, truncated unary code is used for some syntax elements which causes significant increase of context coded bins in the worst case. To improve the throughput of the CABAC in the worst case, a new binarization code called max exponential Golomb code is proposed. By applying the proposed binarization, the number of context coded bins is reduced with almost no coding efficiency loss.

抄録全体を表示

We conducted an experiment on subjective quality evaluation and analyzed the relation between spatio-temporal resolution and subjective quality. Our analysis shows that an effect of spatio-temporal resolution depends on video characteristics.

抄録全体を表示

In this paper, we propose reduction method of the number of filters for block-based adaptation mode on adaptive loop filter by adaptive pattern allocation of filter set according to textural feature of each largest coding unit.

抄録全体を表示

It takes a lot of work and time to take attendance in a lecture that there are many students. So, we suggest the attendance system using iPad in order to reduce teacher's workload.

抄録全体を表示

This paper describes a new Korean mobile learning system for a teacher who has a Korean class. The proposed Korean mobile learning system works in the Bluetooth network and SQLite of database of android device.

抄録全体を表示

We propose a novel real-time color simulator employing (1) 90-degree-hue-rotation method (supports red-green opponent color perception for dichromacy) and (2) simple red-green hue reduction method (enables red-green color deficiency). Subjective test results suggest that the proposed simulator will greatly help mutual understanding of color perception.

抄録全体を表示

We have developed a simple TVML player for Android terminal devices, which can read and play TVML scripts using materials such as character models, motion and sound data on the Internet.

抄録全体を表示

A prototype system of a video streaming multicast system over Wireless LAN has been developed for Android devices. We have also evaluated the performance of the prototype system in the realistic interference environment.

抄録全体を表示

We measured the propagation characteristics of DTTB wave over the Beppu gulf. In this paper, we indicate that the received power fluctuation had a difference in summer and winter, and the degradation of the received power of a certain segment was larger than of all the segments.

抄録全体を表示

The ISDB-T covers Japanese coastal areas. The reception condition on the sea area hitherto has been rarely reported. The data was acquired on the voyage of two training ships. The reception level was almost more than 45 dBuV/m inside of the ISDB-T service area with the omni-directional TV antenna for ship.

抄録全体を表示

Under SFN in digital terrestrial television broadcasting, quality of waves degrades by coupling loop. We calculated equivalent CNR degradation by modeling retransmission inside ship compartment and showed around 6-7dB degradation produces within guard interval.

抄録全体を表示

Usually, electric field strength of a receiving point produces height pattern because of the combination of direct wave and reflected waves. We discussed theoretically height patterns in digital terrestrial television broadcasting, which showed calculated pattern is identical measured one approximately and so on.

抄録全体を表示

This paper discusses the influence of long delay paths mitigation scheme using the two kinds of replica signals that are made from the previous symbol and the re-modulated signal. The proposed scheme achieves less calculation amount than the ordinary scheme using the inverse channel matrix.

抄録全体を表示

It was difficult to measure the signal level below noise level using a field strength measurement equipment. So we proposed the ISDB-T spectrum detection method based on periodicity of scattered pilot to measure super low level ISDB-T signal in the field. This paper describes the field experiment results using the super low level ISDB-T signal detecting device.

抄録全体を表示

We developed a channel filter for the Area Broadcasting, experiments were performed using this device.

抄録全体を表示

Content scrambling system of ISDB-T/ISDB-S employs MULTI2 encryption algorithm. However, any method to update the algorithm is not standardized. We propose an updating method. It uses two algorithms simultaneously, one for current receivers and the other for new receivers. Both receivers can obtain broadcasting services correctly without any disturbance.

抄録全体を表示

Can watch pay-TV in tampering with the IC chip built into B-CAS card fraudulently. By pay-TV broadcasters benefit has been compromised. Issue IC chip manipulation is done to clarify.

抄録全体を表示

We developed a metadata supplementation system for the effective archival and use of the vast amount (tens of thousands of hours) of video of last year's disaster. In this paper, we describe outline functions and system configuration.

抄録全体を表示

According to the recent progress of wireless LAN technologies, short range video transmission using a wireless LAN is getting popular. Since the throughput of wireless LAN is getting smaller when wireless condition become worse, the bit-rate of video streaming system should be changed to adapt this change. In this paper, a method to estimate available bit-rate over the wireless network using both receiver's reports and link layer information of wireless devices is introduced. Our experiment shows that the method significantly improves the robustness of video streaming system using wireless LAN.

抄録全体を表示

The 3-dimensional position data of the camera is necessary in the virtual studio, compositing CG images. Therefore we fixed one end of the wire to the camera and another end of to the constant tension spring, had fundamental experiment. If we set tension in 230gf, the cameraman move freely without feeling it as load, and was able to detect the position within mm order.

抄録全体を表示

We evaluated how much broadcast source, such as a video and a still picture, would carry out image quality improvement with some super-resolving technology. As a result, image quality improved visually by some techniques, and it was shown that this proposal technique is numerically the most effective also in it.

抄録全体を表示

近年,電力不足の懸念から一般照明だけでなく放送業界においても照明の省電力化は大きな課題となっている.このため多方面でLED照明の研究が進められているが,従来の調光装置に直接接続できずLED 化が促進されていないのが現状である.日本テレビでは,既存の調光装置に接続しても滑らかな調光点灯を実現するLED駆動電源装置を考案,LEDフラッドライトとして開発した.また、この製品化においては、価格を安価とするため汎用LED パネルを使用するなど工夫しており,今後のエコ促進に大きく寄与するものである.

抄録全体を表示

The JSPS Grant-in-Aid project, "Brain and Information Science on SHITSUKAN (material perception)", aims an interdisciplinary synergy with neurophysiologists, psychophysicists and engineering area researchers. In this article, we will introduce the objective of this project with current research activities and emerging achievements related to recording, analyzing and reproduction of visual aspect of SHITSUKAN which means material perception or quality sensation.

抄録全体を表示

We propose a new color detection method based on density function of hue variations due to noise. Using our method, we succeed in separation of accurate specular reflection from images including noise. PSNR between correct specular reflection image and separation image is improved 5.68 dB than a conventional method.

抄録全体を表示

We compared freshness estimations of four kinds of vegetable with color and grayscale image stimuli. The results show that color is not critical factor on freshness perception. In addition, we modified the relative luminance contrast between the vegetables and the background, and found that the freshness estimation depends on luminance level and contrast.

抄録全体を表示

Although the physical resolution of digital images is limited by the pixel resolution, one can perceptually judge fine object structures at sub-pixel scales in some occasions. In this study, we attempt to characterize human perception of sub-pixel fineness of hair-like textures.

抄録全体を表示

When synthesizing images with fine objects like hair, we usually use sub-pixel drawing techniques to improve image qualities. In this study, we demonstrate that it is possible to make images look finer by properly adding Gaussian noise.

抄録全体を表示

In the preceding paper, we investigated how noise affects sharpness perception, using white noise and one- and two-dimensional uni-frequency patterns as stimuli. The result showed sharpness of higher frequency stimuli decreased with increased noise, while that of lower frequency stimuli increased at certain levels. From this result, we thought noise reduces sharpness at edges, but can sharpen lower frequency component or texture of the image. In this study low frequency noise or white noise were applied to natural pictures. They were two pictures of tree bark. There was no large difference in their power spectrum, but one had clear edges, and the other did not. The results showed that addition of noise to the picture possessing edges decreased sharpness especially in the case of white noise, whereas noise addition to the picture possessing no edges resulted in sharpness increase at a certain noise level. This suggests that sharpness improvement effect by noise addition depends on whether edges exist or not.

抄録全体を表示

It is known that preference of some photographs is improved by adding noise. We propose "Memory texture" as a reason for the improvement of preference by the noise addition. "Memory texture" is the texture reproduced when we imagine the object. We investigated differences between real texture and memory texture by subjective evaluation of images added noise. Except as to potato image, memory textures were always not smaller than the real texture. Preferred textures were very close to the short-term memory texture in most objects.

抄録全体を表示

In this paper, we derive a novel framework for jointly estimating the reflectance parameters and diffuse texture of a surface of known geometry and illumination from a single image. Unlike previous work, we do not assume the input of multiple images taken under different single light sources, or input of multiple images taken from different viewpoints, or input of multiple images taken under different states of polarization. Under the assumption of spatial homogeneity of reflectance properties, the algorithm searches for the reflectance parameters that minimize the entropy of the diffuse texture image.

抄録全体を表示

We propose a new video segmentation method for the sake of object extraction in video compositing. Our method is based on mean shift segmentation, and has improvements to meet two requirements of segmentation specific to make object extraction effective. We demonstrate the effectiveness of our method by extraction example.

抄録全体を表示