In this paper, we proposc a parallel imagc compression system for high-speed image sensor. The proposed compression system is based on a 4x4-point 2-dimensional DCT using a DA method, zigzag scanning of 4 blocks of the 2-D DCT coefficients and a 1-dimensional Huffman coding. The compression engine is designed with FPGAs, and the hardware complexity is compared with JPEG algorithm. It is found that the proposed compression system require much less hardware, leading to a compact high-speed implementation of the image compression system using parallel processing architecture. The PSNR of the reconstructed image using the proposed encoding method is better than that of JPEG at the region of low compression ratio.

View full abstract

This paper describes a low-noise signal readout technique for CMOS image sensors. The low-noise readout circuits have two-stage architecture with the first gain-stage and the second noise canceling S/H stage. An experimental CMOS sensor with high-gain double-stage noise-canceling circuits is implemented using 0.25μm CMOS technology. Very low-noise level of 34.2 μVrms is achieved.

View full abstract

A future broadcasting system will broadcast along with regular programming a wide variety of program-related information called metadata. Our speech recognition technology is applied to extract information that corresponds to highlights from the commentary of a football game. The proposed method uses additional phoneme models of excitedly uttered words which may be important key words for highlight extraction. The proposed method obtained 14% of key word error reduction, which improved extraction accuracy of scenes making a shoot by 18% in precision and 17% in recall.

View full abstract

This paper proposes a new key frame extraction based on content and context to represent the previous experiences from Life Log video. The context and content of Life Log data are properly applied to extract semantic key frames based on the existence of GPS signal. Our algorithm enhanced satisfied results compared with conventional methods.

View full abstract

This research proposed the facial expression recognition method. In the proposed method, facial organs are detected from complex background by using a real-time template matching and then the facial expression recognition is carried out by using an elastic template matching and eigenspaces.

View full abstract

An entertainment system utilizing a face recognition technology has been developed and reported in this manuscript. This system quickly detects and analyzes face images captured on camera and determines which job category it matches from among the leading-edge jobs flourishing in the Harajuku area.

View full abstract

In the quality evaluation of the coded video, the frame quality is corrected by using criticality information. Criticality information is made to cluster by the k-mean clustering method, and the frame quality is corrected by the representative point.

View full abstract

In this paper, we describe the objective quality evaluation model for moving picture. Our proposed model is the Reduced Reference model that uses several features derived from reference video. These features are texture features, temporal degradation and flat/detail area information.

View full abstract

JSEG is a powerful segmentation algorithm which consists of initial labeling stage and spatial segmentation stage. The authors have proposed two algorithms based on k-means color clustering and Fuzzy c-means color clustering for initial labeling. This paper verifies goodness of those algorithms and proposes a modified algorithm.

View full abstract

To make meta database for MPEG-7 Annotation, we propose new clustering technique. At first, to detect scene change point, and make clusters of scenes, called "cuts". Next, make clusters of cuts. We suggest new method "Center Gravity of Outline" to detect scene change point and make clusters.

View full abstract

We propose a function-oriented classification of web images. We defined nine categories of images taking into account of their functions, and classified web images by using SVM in tree structured way. We show the page summarization by our new categories of WWW images.

View full abstract

As one of features, we have an interest in human skin as physical feature of human. We aim to distinguish ages by calculating the fractal dimension as a feature of a skin surface image. The effectiveness of feature extraction is examined by experimenting with skin surface images of two ages.

View full abstract

The culture of the sea bream is performed by Kinki University. It occurs often that the shape of sea bream fries are to be abnormal. It is necessary to automate the selection of normal and abnormal ones. Then, the pattern recognition is used in this research to select them automatically.

View full abstract

This paper investigates fundamental characteristics of mobile camera-based Chinese character recognition. For two kinds of typical OCR algorithms, we verify the influence to recognition accuracy of light source, font type and character resolution. Experimental results using Chinese restaurant menus show that contour features are more robust for environmental changes than skeleton features.

View full abstract

When taking a picture of animation, focusing on and tracking the human may be work of professional. In this paper, as a preprocessing to automate such work we propose and examine a face detection algorithm using template matching with skin color extraction.

View full abstract

It is difficult to identify of low resolution number on carlicense plate recorded on surveillance. A method for identifying low resolution number on carlicense plate is proposed. Identification is performed by using the distribution of gray pixel and geometrical feature like upper and lower, right and left symmetry.

View full abstract

We can record the appearance of streets using in-vehicle small camera. Same scenery is repeated in car video so the change point is very important. To detect the change point, first we have to look for two images taken at the same point and compare them. Using the location information and image features, we can find the same points.

View full abstract

A system for video summarization in a ubiquitous environment is presented. Data from pressure-based floor sensors are clustered to segment footsteps of different persons. Video handover has been implemented to retrieve a continuous video showing a person moving in the environment. Several methods for extracting key frames from the resulting video sequences have been implemented, and evaluated by experiments. It was found that most of the key frames the human subjects desire to see could be retrieved using an adaptive algorithm based on camera changes and the number of footsteps within the view of the same camera. The system consists of a graphical user interface that can be used to retrieve video summaries interactively using simple queries.

View full abstract

A lot of alarms are generated at the same time when the equipment of terrestrial Digital TV transmitter has malfunction. Therefore, it is difficult to specify the equipment that causes trouble. This report is for a comprehensive Data-logging system that is developed to specify the broken equipment.

View full abstract

We developed BML browser for broadcasting system based on home servers. Five new functions, playback, search, segment table lookup, program information lookup and status lookup, are implemented to this browser.

View full abstract

A metadata production system has been developed to create and edit segment metadata of the semantic level by combining various methods of metadata extraction. The system is composed of functional modules for production, extraction, accumulation, editing, and utilization of metadata. This paper describes the framework for metadata production and the experimental metadata production system.

View full abstract

It is very difficult for an amateur to produce a sophisticated broadcast program, because the television program manufacturing equipments are very expensive and need a lot of professionals. Here, we discuss the program production device that even an amateur can master easily on a personal computer.

View full abstract

This system can measure the long term fluctuation of the conditions such as receiving power, carrier to noise ratio, bit error rate and so on and logged. It also has remote monitoring system to send the reception condition data from remote site.

View full abstract

For a baseband transmission of MPEG-2 TS streams over an optical fiber, we developed a experimental equipment. To validate the fundamental functions experimentally, MPEG streams were multiplexed to transmit over an optical fiber. That shows the possibility of re-broadcasting with digital baseband MPEG-2 TS over an optical fiber.

View full abstract

In OFDM transmission, performance is degraded by the existence of time domain impulse noise because its influence spreads over the whole carriers by the DFT operation in the receiver. Especially influence of the impulse noise whose amplitude is large. In this paper, the scheme of improving influence of impulse noise is proposed. In the proposed scheme, the symbols after decision are re-modulated to generate a time domain signal to replace the samples that are degraded by the impulse noise in the time domain. The time domain signal where the influence of the impulse noise is reduced is demodulated again.

View full abstract

Illegal copies are critical issues in content business. It is necessary to develop a new technology to deter them. In this paper, experiments are made for identifying content with two methods: extracting characteristic information of video, and phase only correlation. The result shows that combination of these methods is effective.

View full abstract

We have developed a new experimental feature that enables us to perform a phase only correlation analysis on the characteristic information analysis system for video content identification. This improves the identification rate in single characteristic information analysis without raising the computation cost much.

View full abstract

Recently, ion-beam irradiation developed as a new liquid crystal (LC) alignment technique without contact with alignment layer. However, alignment mechanism of LC by the ion-beam irradiation has not been clarified. In this paper, to clarify the alignment mechanism, the correlation between polymer alignment and LC alignment condition was analyzed.

View full abstract

As the evaluation method of surface order parameter of the liquid crystal on a polymer surface, the phase transition droplets method and slanting incident IR-absorbance method have been used. In this paper, reliability of measurement of surface order parameter was discussed.

View full abstract

The surface polar anchoring strength is the most crucial element which governs the response property of VA-mode LCD. We devised a new measurement method for the surface polar anchoring strength and the elastic constant using a hybrid alignment liquid crystal cell. We measured the angle dependency of the phase difference in the direction perpendicular to the rubbing direction of HAN cells and demonstrated that these parameters can be measured with a high degree of accuracy.

View full abstract

We have developed an improved new scaling algorithm, which we call "Second generation ASPECT (Advanced Scaling with Preserving Extracted Contour's Transition)". ASPECT2 is able to improve the Edge to sharpness without overshoot or undershoot

View full abstract

We have studied Feedforward driving with image compression. It achieved the response time improvement of LCD displays and the memory reduction. In this report we discuss the method for preventing the deterioration of response time improvement performance because of memory reduction when a particular pattern of image data inputs.

View full abstract

We have made pictures for evaluating multi-primary monitors. These pictures include color data out of a color gamut of a three-primary-color LCD using three-color LEDs. Using these pictures, we have compared color reproducibility of both a monitor output and a printing output, and had confirmed these monitors may be adapted for soft proofing.

View full abstract

We have been developing the new Fresnel lens screen suitable for a wide-angle projection optics. It fundamentally consists of a refraction and a total internal reflection prism. In this paper, we summarize the design concept of hybrid Fresnel lens screen and expected performance for ultra-thin rear projector.

View full abstract

We have developed a new projection optics with the exceptionally wide field angle of 160 degrees and the short projection distance of 125 mm with a 62 inch screen. In this report we measured quantitatively image resolution for the projection optics and compared the measured values with the simulation values.

View full abstract

We suggest "Environmental Display". The "Environmental Display" is a system which provides territories to display information to each and every one of users. Each territory follows each user, and he/she can always check necessary information. In this thesis, we make prototype system with PCs and projectors, and evaluate following accuracy.

View full abstract

Normal maps play an important role in realistic 3D image rendering to express pseudo roughness of the surface with small amount of polygon data. In this paper, an efficient normal map compression algorithm is proposed based on vector quantization. Using the unity condition of normal maps, compression ratio has been made much better than conventional approaches. In addition, the encoding time has been made reasonable by considering the unity condition and the distribution of the data. The compression algorithm developed in this paper is also compatible with hardware accelerators.

View full abstract

In the process of encoding MPEG2, entropy is measured before and after giving a binary encoding, and the indicator of the compression efficiency and the improvement of variable-length encoding of the use with MPEG2 is verified.

View full abstract

We previously proposed new motion-compensated prediction based on adaptive interpolation for video coding. In each block, a motion vector is detected with integer-pel accuracy, while a predicted image is produced using one of adaptive interpolation filters. In this paper, we introduce the overlapped block motion compensation technique into the prediction method to reduce blocking artifacts of the predicted image.

View full abstract

We previously proposed a novel lossless coding scheme in which several coding parameters are optimized for each image. To extend the scheme to near-lossless coding, in this paper, a new cost function suitable for coarse quantization of prediction errors is introduced. Simulation results indicate that the proposed scheme outperforms the JPEG-LS standard in both lossless and near-lossless coding.

View full abstract

We have studied a method to reduce video coding noise such as blocking artifact and mosquito noise in compressed video using fuzzy algorithm. A practical postfilter is considered to implement the studied method into hardware. It performs good to reduce the noise and is available for many purposes.

View full abstract

This paper examines how to reduce the redundancy between signals that exist in the luminance and the chrominance in order to improve compressibility. We propose the method of predicting the chrominance from the luminance by conversion using a few coefficients. The simulation experiment shows that the proposed method is twice as effective as the conventional method at the low bit rate.

View full abstract