We developed a data communication system for protection and disaster relief operations using TV white space for disaster rescue-robots which worked in a prohibited entry area. This paper reports experimental results of interference with a wireless microphone, and propagation characteristics in this system.

View full abstract

A standard of DVB-S.2 defines roll-off over 20%, however a standard of DVB-S.2 Extension, under consideration, defines roll-off 5%. We tested SNG transmission with new technology, include 5% roll of and 32APSK. In this report, we describe the result of the test and DVB-S.2 Extension is quite effective.

View full abstract

[in Japanese]

View full abstract

We proposed KANSEI evaluation method to evaluate impressions that viewers receive from super high resolution images. We could confirm to improve the impressions such as beauty or the feel of material of super high resolution images in comparison with full high-definition images.

View full abstract

Background replacement using a chromatic adaptation transform fits a color of a foreground image to an illuminant of a landscape used as new background image, and can enhance realism of video chatting. In this paper, to improve the realism, we examine an illuminants estimation method of the background replacement.

View full abstract

A method for differentiating between computer generated and digital camera images based on wavelet analysis is presented. This scheme models the marginal distribution of a wavelet coefficients using a Gaussian mixture. However, since these densities have more sharp zero peaks and heavy tails, such a mixture model is difficult to stably estimate. In this paper we propose a robust technique for classifying computer generated and photographic images based on a generalized Laplacian distribution modeling.

View full abstract

Chromosome analysis is one of useful crews to diagnose diseases and has been carried out manually by specialists. The specialists are therefore required expert skills and rich experience after a long period training, causing the shortage of the number of such specialists. Chromosome analysis with using the pattern recognition techniques can be a great aid to solve the problem of this shortage. Previous systems of digital image analysis by pattern matching methods can operate only limited processes automatically because chromosome images have a lot of variable image properties to identify, such as a size, linearity and color stained. It is well known that G-band pattern of each chromosome in karyotype is unique and key to find whether normal or not. In this study, we therefore propose an algorism to compare a luminance profile of the G-band pattern of a chromosome image and of the ISCN ideogram of each chromosome.

View full abstract

In the pupil-corneal reflection-based gaze detection, the glasses reflection of the near-infrared light source tends to be misdetected as the pupil. In our previous study, we proposed the pupil searching and tracking methods using a constant interpupillary distance as the constraint. However, there were some cases that the pupils are not detected because the distance actually changes. To improve this, in the present study, the method to renew the distance using its histogram is proposed.

View full abstract

A method of pointing device implementation for mobile cameras using Radon transform is proposed. Moving of the mobile camera can be chased by its video image. We move a mouse cursor on an arbitrary image in the display by the mobile camera movement. The Radon transform is effective in a description for rotation and translation of the scenes.

View full abstract

We have proposed an application named "Focus UI" which recognizes a real object and connects to a related digital data. In order to develop the application, we propose a method for extracting an area based feature which performs object recognition of 99% accuracy in speed-of-response 0.5 second.

View full abstract

In this study, we investigated a variation and temporal change of skin temperature of cheeks when subjects rest and watching TV-comedy program as emotion-eliciting stimuli. We acquired the skin temperature using thermography in 30fps. Experimental results revealed that the cheek temperature rises when subjects having emotional responses in most cases.

View full abstract

In this paper, we propose a classification method of human actions from first-person videos. Our proposed method detects visually salient objects and integrates saliency value and object likelihood into feature vectors. The experimental result shows that the proposed method can classify first-person actions more accurate than a previous method by 6.4 percentage points.

View full abstract

We propose a new method to interface with 3D-CG model by using augmented reality (AR). By detecting fingers of a user with a Kinect sensor, he can deform the CG model as if he touches a clay model. The AR technology can enhance the reality of manipulation of the model.

View full abstract

Multi-screen Broadcasting Study Group examines the broadcasting Service of synchronizing second screen contents with aired TV programs.

View full abstract

[in Japanese]

View full abstract

The rapid increase in digital devices and the content such as digital photo and song have made it possible for many users to create music slideshows. However the editing process to select songs and photos requires much intensive effort. This paper proposes music slideshow generator which enables to select song which suits context of photos based on relevance score.

View full abstract

This paper proposes an area-efficient low-power column-distributed ramp signal generator without any power-hungry amplifiers. This consists of miniaturized column-wise ramp generators connected together by metal wire, reducing eventual signal mismatch. Simulation results show less than 1LSB gain error of the ramp signal is attained only in the area of 2.4mm^2.

View full abstract

We designed a new column parallel ADC to be mounted on the Super Hi-Vision imaging sensors. This ADC is a pipelined ADC that consists of a cyclic ADC and a successive approximation (SA) ADC, and this structure provides fast speed conversion and lower power consumption. Using simulators, we confirmed this ADC functions correctly and its DNL is within 1 LSB.

View full abstract

A CMOS image sensor with a low noise transistor is developed. The proposed imager has a low vertical fixed pattern noise (vFPN) and small random telegraph signal (RTS) noise because of the noise reduction effect of the native transistor as an in-pixel source follower amplifier.

View full abstract

By using nanofabricated 0.11 μm CMOS image sensor process, prototyped Super Hi-Vision image sensor indicated a conversion gain of 112 μV/e^- and a sensitivity of 2.25 V/lx-s, respectively. Furthermore, the random noise was 2.1e^^-_<rms> with a column amplifier gain of 17.

View full abstract

We have developed a Super Hi-Vision camera with spectral sensitivity characteristics corresponding to the wide-color-gamut colorimetric system. We measured color reproducibility and the camera practically provides acceptable level of color reproduction performance including high saturation color.

View full abstract

The device simulation with considering propagation of electromagnetic field induced by gate electrodes is realized by extension of Maxwell's equations and finite difference time domain (FDTD) method. It is shown that floating diffusion amp (FDA) noise induced by electromagnetic wave strongly depends on the substrate thickness.

View full abstract

SECOND-MVD (Synthesis Error COmpeNsateD Multiview Video plus Depth) is an alternative 3D format, where images at some viewpoints remain original, and the others are converted to a novel format. Depth stabilization improves time domain depth consistency in MVD format. We survey effect of depth stabilization on performance of SECOND-MVD.

View full abstract

Recently, the coding technology of multi-view video plus depth has been developed to implement 3D video applications such as Free Viewpoints Television. Because depth maps tend to have smooth value within an object and have strong edges between objects, it is efficient to encode depth maps by a region-oriented manner. In this paper, we introduce a segmentation method based on Voronoi diagrams to reduce the amount of information on the region shapes. Moreover, we propose a plane approximation to the depth values in each region.

View full abstract

Recently, with the development of the 3D technology, it is commonly used to see and hear 3D contents through 3-D TV and PC for a long time among at home. So we focused on the sitting posture and the parallax of 3D contents as fundamental researches about this and researched it. As a result, a correlation was seen in both sides.

View full abstract

We have conducted objective evaluation experiments on depth perception and image sharpness. The results showed that the observers could perceive the image sharpness, however, the depth perception didn't accord with the depth positions of the 3D target. Therefore, we have suggested the optimum depth range of reproduced Integral 3D images.

View full abstract

We developed a prototype of Integral 3D display with multi-projection system using Fresnel lens and lens array to be able to expand the viewing zone in the horizontal and vertical direction. The projection image on lens array is distorted by the projection from the oblique direction and Fresnel lens. We generated 3D images by compensating the distortion using projective transformation and affine transformation. We confirmed that the viewing zone was expanded up to 2.47 times in the horizontal direction and up to 1.99 times in the vertical direction compared to the viewing zone by using one projector.

View full abstract

In this paper, we propose a method of accurate detection of vegetation areas in the urban areas using high-resolution satellite images. Using multiple thresholds, we devised a method for reducing influences of noises or shadows and increasing detection accuracy of vegetation areas.

View full abstract

This paper proposes a video summarization support system by interactive evolutionary computation. The system evaluates a solution on the basis of qualitative (user preference) and quantitative (redundancy and inappropriateness of video length) criteria. The system works well without depending on video category because of avoiding the use of features depending on the category.

View full abstract

In this paper, we propose a color transformation method based on color theme that can i) control the ratios of assigned colors while ii) maintaining the spatial coherence of regions. We introduce a labeling approach to formulate a color assignment to each pixel. It is solved using a cost volume filtering with an adaptive offset for each color label to ensure the minimum appearance ratio of the color throughout the picture. Experimental results show that our approach can control of the color ratios while preserving the coherence of the regions.

View full abstract

This paper describes a method for extracting keywords which help for users to know why the item is recommended in recommender systems. We conducted evaluation experiments, and confirmed the effectiveness of our proposed method.

View full abstract

We propose an image attribute classification method that employs a k-nearest neighbor (k-NN) classification. The proposed method employs color CoHOG to represent each image and classifies by using large margin discriminative hashing. Experiments indicate that the proposed method exhibits high accurate classification performance.

View full abstract

In the previous work, we proposed a lossless coding method which dynamically models probability density function (PDF) of image values pel-by-pel. In this method, the PDF model is defined as sum of the generalized Gaussian functions (GGFs) whose peak positions are given by examples collected from causal neighbors. However, the shapes of the GGFs were fixed for each image. In this paper, we introduce a technique of context modeling to adaptively control the respective shapes.

View full abstract

We previously proposed a lossy coding scheme based on recursive-type intra prediction and discrete sine transform (DST) for monochrome still images. In this paper, we extend the scheme to RGB 4:4:4 format color images. In order to exploit intra- and inter-color correlations, the already encoded color signals are simultaneously used in the prediction process.

View full abstract

A standardization work of High Efficiency Video Coding (HEVC) version 1 has been finished and its extensions to a non-4:2:0 chroma format is currently studied. In this paper, we propose a coding method based on a blockwise un-correlation of chroma space for intra/inter residual signals. A transformation matrix for the color-space un-correlation is derived from the coded sample by the simplified singular-value-decomposition (SVD). It is confirmed that a bitrate reduction is 23.4% for RGB 4:4:4 materials and a bitrate reduction is 4.5% for YUV 4:4:4 materials.

View full abstract

The averaging attack, which is one of collusion attack methods, is very forbidding. We proposed the seam carving technique based geometric transforming scheme by considering ROI. This proposed method deteriorates an average attacked image quality. Experimental results show that good subjectivity image quality of transforming images.

View full abstract

We have proposed a language model adaptation method for captioning disaster broadcasts. In this paper, we develop a prototype system to collect training texts for the adaptation. Through experiments on speech recognition, the effectiveness of our system was evaluated.

View full abstract

We developed metadata supplementation system for the effective archival video of the East Japan great earthquake disaster. We evaluated the metadata detection accuracy of the system and estimated the metadata production time by using the video more than 40 hours it was taken during the earthquake.

View full abstract

A project is ongoing at SMPTE to enhance applications of UMID, the SMPTE standard globally unique AV material identifier. While already identified the UMID application principles which are to be reflected to upcoming revised SMPTE RP 205, how to realize the UMID resolution protocol to convert a given UMID into its URL is under intensive study. In this paper, we report the latest status of the project as well as the feasibility study of DNS (Domain Name System) to be used as a basis of the UMID resolution protocol implementations.

View full abstract

The "second screen services" related to TV programs enrich broadcasting service. This paper describes about our trial services using the AR application on smartphone collaborate with TV programs. The AR application can enhance TV viewers' experience by providing 3D-CG-object related the TV program.

View full abstract

A high sensitivity camera which used 2/3-type single chip color HbCMOS sensor was developed. The starlit sky was able to be pictured and the clear image was obtained. As compared with the high sensitivity camera which uses EM-CCD, the performance and a future possibility to the high sensitivity of this camera are shown.

View full abstract

In this study, we successfully fabricated a photosensor using red (R)- and green (G)-sensitive organic photoconductive films with In-Ga-Zn-O thin-film transistor readout circuits. The R- and G-sensitive layers were separated by 10-μm-thick interlayer insulator. All fabrication process was implemented at a temperature of 150℃ or less.

View full abstract

High-dose radiation destroys cancer cells in radiation therapy. Verification imaging is performed to adjust the patient's position to precisely target the tumor. We have applied a refraction contrast imaging technique to verification imaging to improve the accuracy of radiation treatment. The edge enhancement effect in refraction contrast imaging can produces higher contrast images than those in conventional imaging. In this study, rat head images were obtained by refraction contrast imaging and conventional imaging for the evaluation of image quality improvement.

View full abstract

A new active-matrix Spindt-type field emitter array for a FEA-HARP (field emitter array-high-gain avalanche rushing amorphous photoconductor) image sensor was designed. Simulation and experimental results showed that the new active-matrix drive circuit can shorten the response time and can potentially meet the HDTV standard.

View full abstract

In order to achieve an ultra high speed imaging toward 1G frames per second, the novel structure called "Multi-Collection-Gate Backside-Illuminated image sensor" has been proposed. Based on the structure, this paper reports a preliminary pixel design and simulation results of "Nonosecond-resolution image sensor (NanoSIS)".

View full abstract

We developed an implantable CMOS imaging device for detecting weak green fluorescence in a living body such as a mouse brain. By coating an excitation filter on LEDs, we achieved excitation light transmittance of 10-5 through a fluorescent filter on the image sensor.

View full abstract

Computational photography has redefined the possibilities of conventional photography. By mixing modified hardware with software, new effects can be obtained. In this paper we present a theoretical framework for a computational camera, based on coded apertures, that is able to multiplex images of a scene and simultaneously acquire them. As an application, we show a system that can capture images with different exposure values, a feature often useful in high dynamic range imaging, for example. The framework is in no way limited to the acquisition of images with varying exposure values. It is adaptable to different types of filters, like color filters. The advantages of this system are the low cost of implementation, ease of adaption to different conditions, and the use of full sensor resolution. However, the need and difficulties of the decoding process can create image artifacts.

View full abstract

A CMOS image sensor with stacked photodiodes has been developed to observe fluorescence in lensless digital ELISA (enzyme-linked immunosorbent assay) system. The sensor has been fabricated by using 0.18μm mixed signal CMOS process. Sensitivities of the both stacked photodiodes indicate that they can distinguish the fluorescence from excitation light which cannot be completely eliminated by an on-chip fluorescent filter.

View full abstract

In this paper, we propose the methods of dynamic range expansion and motion blur suppression using the image sensor whose focal plane is divided into several blocks.

View full abstract

In this paper, we propose the image sensor that can control reset and readout timing of equally-spaced pixels in a two-dimensional direction. We explain the structure of our proposed image sensor and consider the subsequent signal processing.

View full abstract