This paper proposes computer aided diagnosis (CAD) system that uses the image recognition technique based on the higher-order local auto-correlation features (HLAC). The computational simulation implied that the proposed system could detect the gastric biopsy image containing cancer.

抄録全体を表示

In this study, we try to read printed characters on the book. This paper describes the method to segment a sentence to characters and search a pronunciation of the character by the dictionary.

抄録全体を表示

We produced the technique for recognition to gestured handwriting characters by the secure system equipped with a touch-sensitive panel on the front of a door, a gate, and so on. It proposes to employ the touch-sensive panel to acquire fingerprints, and to draw a secret word with the fingertip. This paper describes the pattern matching method to detect the position of the finger-tip at high speed.

抄録全体を表示

In this paper, we propose a method for increasing the ratio of correct corresponding points in the different view point building image retrieval using SIFT algorithm. We create canonical views and match respective vertical lines based on geometrical constraints. The method improves the ratio of matching accuracy by 8%.

抄録全体を表示

We propose the new fitness system which combined more than one piece of fitness equipments in the virtual reality space. We analyzed the change of the pulse of the examinees and their subjective evaluation during the exercise using the fitness equipments that interlocked with various VR space.

抄録全体を表示

The focus of this research composes a vocal recognition method without voice recognition. The principal aim is establishment of the method of analyzing the lip movement when speaking. And, we are proposing the technique of the utterance recognition based on the correlation of the power spectrum obtained from Fourier translation.

抄録全体を表示

In the previous study, we have developed the small scale device which can measure eye movement with one personal computer. Improving this device, much more portable and smaller eye movement analyzer with GPS receiver has been developed. We analyzed the eye movement during walking up the sidewalk.

抄録全体を表示

The focus of this study is to construct utterance recognition method without voice recognition. We have been already clarified that there vowels of British English were discriminated by Fourier Transformation of the lip movement. Here, we expanded the study to the lip movement when speaking the other vowels of British English.

抄録全体を表示

There are various input devices. Nevertheless, input methods are not unified. A user has to adapt an input method to each device. We propose a system which adapts itself to a user. The system has cameras, and recognizes the user's operation of his or her own mobile phone which a user carries in daily life. We got fine experimental results.

抄録全体を表示

For line-of-sight detection, we have detected the pupil and the corneal reflection of the light sources using the two light sources and the image difference method. However, detection was difficult under highly illuminated environments. In this study, we improved the method by additionally obtaining no-illumination images and proposing alternative image processing.

抄録全体を表示

We propose the hand gesture interface to operate a wearable PC in outdoor using image processing. To recognize the hand movement, hand area is expanded from the one block of hand. Next, the shape of hand is recognized by fmger position. And, the shape is acquired in the time series.

抄録全体を表示

We propose a countermeasure to stop re-shooting movies on the screens through digital cams. Conventional measures using digital watermark embed content ID into pictures and videos so that one can identify place and time where content is re-shot. Conventional method, however, cannot control re-shooting by digital cams. The proposed measure can stop re-shooting by destroying shot content using invisible signals which add noise to content shot through CCD/CMOS devices. Testing of a prototype system implemented in 100-inch screen demonstrated the validity of the proposed measure.

抄録全体を表示

We measured every pixel signal of every frame from HDTV broadcast signals continuously during one week. Average picture level and CIE chromaticity values (u', v' and a^*, b^*) were calculated from these data. The results indicate that the average APL is 44.8%. Color gamut requirements for broadcast signals were discussed from 2D chromaticity histograms.

抄録全体を表示

We investigated the reduction of disturbing reflection on FPD-TV screens by inclinating the screen. The results indicate that the probability of reflected glare from luminaires reduced from 37% to 5% for 80 inch TV by inclinating the screen up to 10 degree.

抄録全体を表示

We evaluated image quality of seven HDTV LCDs from different manufactures by using ranking method. The results showed the subjective ranking of image quality of LCDs strongly depended on the characteristics of the displayed image contents. Relationship between the results of image quality evaluation and the contents of images were discussed.

抄録全体を表示

This purpose of this study is to discuss the perceptual phenomenon of movement object on LCD on the basis of signal processing of visual perception. The simulation and the psychophysical experiment shows the motion blur width depends on velocity of movement object, not that of visual angle.

抄録全体を表示

In this paper, we describe a new super resolution method which reduces the number of iterations. We simplify the image degradation model by using non-iterative preprocessing algorithm. Simulation results confirm the effectiveness of our method and demonstrate its superiority to previous method.

抄録全体を表示

In this paper, we propose a new method to improve moving picture quality by Sub-Field relocation by using motion vector. We confirmed that the proposal method can reduce the image noise known as dynamic false contour and motion blur.

抄録全体を表示

A hybrid write- and erase- addressing technique is employed for the ultra-high-resolution PDP. Gray level expression is successfully achieved when the scan pulse widths for write addressing and erase addressing are 1.0μs and 0.5μs, respectively. The address voltage is 55V.

抄録全体を表示

We analyzed the polarization shift inside twisted nematic liquid crystal display (TN-LCD) and clarified the S-shaped polarization shift on Poincare sphere in black state was important to realize low-voltage driving TN-LCD. We realized high contrast ratio of 150000:1 with low driving voltage of 2.5V by optimization of optical compensation films.

抄録全体を表示

We conducted the subjectivity evaluation experiment for the "Color Uniformity" with 40 inches LED backlight, and subjectivity evaluation value got a result to rise so that an area and the chroma of the uneven coloring were large. Furthermore, we studied the techniques to extract the indexes considered the hue influence.

抄録全体を表示

We have developed a new projector with the fourth new liquid crystal panel that controls the yellow light independently in UHP lamp. In this paper, we present a novel signal processing technology for the projector. This Technology controls the volume of yellow light based on brightness and chromaticity analyses. As a result, this projector achieves both high brightness and vivid color reproduction.

抄録全体を表示

In this paper, we propose a method to reduce motion blur and random noise by using two coded exposure patterns. The method was implemented on a real time imaging system using an image sensor with iterative charging function and we evaluated the effect of image sharpening.

抄録全体を表示

This paper propose color-imaging scheme with random-arrangement-type RGB color filter array for single image sensor. There are two advantages in the use of random CFA. The one is able to set the ratio of the RGB color filters arbitrarily. The other is able to suppress the influence of the elasing artifacts.

抄録全体を表示

Micro-angiography with spatial resolution in the micrometer range was carried out to depict vascular responses of the pulmonary arteries in rats using a real-time imaging system and a synchrotron radiation source at SPring-8. In imaging experiments, constriction and dilatation of small arteries were visualized in response to drugs.

抄録全体を表示

The conventional wiper performance is the evaluation which emphasized only the wipe range. It is necessary to ensure the visibility for the performance evaluation of the wiper. We propose techniques for measuring a wiper based on the performance of wiping raindrops and shielding effect by the wiper.

抄録全体を表示

Toward an ultraminiature high-performance microphone for the next generation, we are studying a silicon microphone. With the aim of low-voltage operation for this microphone, we developed a charge storage technique using soft X-rays. Experimental results show that internal inorganic dielectric film storages and keeps electric charges, which indicates that this dielectric film becomes electrets. This technique shows promise as a low-voltage operation silicon microphone.

抄録全体を表示

On the basis of the analytical transfer model we developed, we investigated the suitable magnetic properties of garnet film for transfer from magnetic tape. Signal of recorded wavelength of 0.77μm transferred clearly from tape to garnet film with appropriate magnetic properties. We showed a relationship of readout signal and film thickness of garnet films.

抄録全体を表示

In this study, we report on the magneto-optical characterization and the magnetoresistance properties of perpendicular magnetic tunnel junction for submicron sized magneto-optical light modulation device.

抄録全体を表示

To realize ultra-high difmition spatial light modulator (SLM), we are developing a new magneto-optical SLM driven by spin transfer switching. In this study, we observed 1 μm size of spin valve pillar using magneto-optical microscope and succeeded in getting clear image of magnetization reversal.

抄録全体を表示

Recently, natural secret image hiding schemes for two binary halftone share images have been proposed. In order to generate images of high visual quality, we introduce swap operations into the conventional method which only uses toggle operations. Simulation results show that the method with swap operations can achieve higher image quality.

抄録全体を表示

In this paper, since past degradation caused by flicker was improved by the one-dimensional histogram analysis method, it report. Since processing time is very fast, this technique has high practicality.

抄録全体を表示

In this paper, deterioration in the image by the scratch mark was able to be improved greatly by the interpolation method by making the straight line easy to detect from the difference picture. There is a practicality because it is a very high repair rate and high-speed processing time.

抄録全体を表示

A new method of compression artifact reduction is proposed. This method applies kernel regression to compressed images, which approximates the luminance pattern by smooth surface. The coordinate of the surface is rotated along the edge direction and the surface is defined appropriately for the edge shape.

抄録全体を表示

Ultra-high resolution (8K) video is expected as one of the next generation video formats. In order to provide consumer service with 8K video, high performance video coding technology is strongly required. In this work, we evaluated the subjective picture quality of 8K video coded by H.264 technology, and the coding performance of advanced coding tools was also evaluated. From these results, we discussed future possibility of 8K video coding.

抄録全体を表示

We have developed a software encoder of the broadly used H.264. In order to achieve fast encoding while maintaining compression efficiency, it uses various speed-up techniques and subjective visual quality driven encoder optimization. In this paper, we report some of the core methods and the performance of this encoder.

抄録全体を表示

In this paper, a new motion vector coding method, which enhances coding efficiency of the H.264/AVC based coding, is proposed. The proposed approach utilizes the knowledge of predicted motion vectors in order to reduce bits required to encode motion vector differences. The experiments show that the proposed method's effectiveness.

抄録全体を表示

We propose the new method of high dynamic range image composition. The proposed method can achieve composition by three steps. It first constructs intensity histograms of reference images. Then second, it extends histogram by piecing particular parts of reference histograms. Finally, it converts pixel values according to extended histogram. The performance of the proposed method is efficient in spite of its simplicity.

抄録全体を表示

We propose an image binarization method for preserving local contrasts between neighboring bright and dark regions. We propagate the information of the sign and intensity of gradient of principal edges to the whole image by using iterative cross bilateral filter (ICBF), and then binarize it with a simple thresholding. Experimental results on synthetic and real images show the effectiveness of the proposed binarization method.

抄録全体を表示

Conventional photometric stereo cannot recover correct surface orientation when interference due to specular reflection and/or self-shadowing is observed. To overcome this problem, appropriate light sources which eliminate the interference are selected pel-by-pel.

抄録全体を表示

This paper describes influence of combining moving pictures in Image High Resolutionalization Using Multi-Point Local Search. We analyze output images in order to evaluate diagonally moving pictures.

抄録全体を表示

An inpainting system for largely occluded region is proposed. The inpainting utilizes SIFT feature point detection and a geometrical transformation of the second view image.

抄録全体を表示

A comparison of obfuscation methods for watermarking system with disclosed detector is qualitatively done. Essential numbers needed in the detector software are encrypted by decoding keys are obtained from results of the software execution on the way. Based on evaluating the decoding time, we can make the watermark detector having sufficient computational amount.

抄録全体を表示

A medical diagnosis based on image sequences requires to protect their tampering and spoofing from malicious intruders. This paper describes a method for a tamper-proof procedure by using non-referred lossless data hiding and a prevention of spoofing by using the hybrid encryption scheme.

抄録全体を表示

This paper proposed an automatic image collection method through the world wide web. The proposed method calculates the similarity between image features which are retrieved by the Google image search and obtains proper images recursively. Simulation results confirmed the proposed method could search many proper images more than the Google.

抄録全体を表示

In this paper, we propose the technique that advanced face detection from a photograph and apply the best shot in the group photograph.

抄録全体を表示