Japan started a digital data broadcast using the BS (Broadcasting Satellite) in December 2000. For digital data encoding representation, it uses XML (Extensible Markup Language) based information representation which is called BML (Broadcast Markup Language). The aim of this language is seamless transforming to HTML (Hyper Text Markup Language), which is familiar to use in WWW contents and also extending functionality for broadcasting requirements, such as media synchronization, event handling and navigation. This paper shows the specification of this language and technical issues.

View full abstract

This paper presents a new image estimation method for removing the image of eyeglasses from facial images of people wearing eyeglasses. Wearing eyeglasses can change one's appearance. It is necessary to estimate eyeglassless facial images from eyeglassed facial images for evaluating impartially characteristics of both eyeglassed and eyeglassless faces. When we distort an eyeglassed facial image using Wire Frame Model into a facial image showing a different expression, the eyeglass frame is deformed. Therefore we need to extract and remove the eyeglass frame region before the distorting. There are some methods for estimating eyeglassless s facial images by interpolating gray levels in the eyeglass frame region. However these results contain unnatural regions and eyeglass shadows. Our method uses basis vectors, calculated by Principal Component Analysis to obtain eyeglassless facial images, which better capture the characteristics of “eyeglassless faces”. The estimated faces appear natural and we numerically evaluate them using their similarity to eyeglassless facial images.

View full abstract

Falling snowflakes were photographed simultaneously by two CCD cameras having different shutter speeds. The high speed camera was used for determining the shape and the low speed one to measure the motion. The shape of a snowflake and its falling motion are analyzed using continuously photographed image frames. The falling motion of a snowflake is calculated by tracing its position in subsequent frames. Relationships between shape and motion parameters were analyzed. Data indicate that tilt angle and falling direction were perpendicular to each other. It is considered that snowflakes encounter more resistance from air toward the upper direction during their fall.

View full abstract

We propose a new detail discrimination method using the differential stroke which is extracted automatically to take the maximum stroke distance among whole corresponding strokes from two reference patterns. The method forms each common pattern by eliminating the differential stroke from each reference pattern, then obtains each remained stroke by eliminating the partial pattern corresponding to the common pattern from an input pattern, next performs discrimination using each differential stroke and each remained stroke. The method possesses an advantage that there is no load to edit the dictionary if the new similar character pattern appears because the differential stroke is extracted automatically when the similar character patterns are presented. In the discrimination experiment for Japanese Industry Standard level-2 Kanji character pattern written in correct order-correct number and the Kanji character pattern in constant use written in the square style, the proposed method improves the correct recognition rate to both experiments, and its effectiveness is confirmed.

View full abstract

As fundamental similarity measure for robust image registration for change in brightness, the rank correlation coefficient is investigated with real image registration problems. Sampled rank correlation, furthermore, is proposed for realization of robustness for occlusion. The rank correlation is a statistic which is known to be robust for large change in brightness in scenes due to illumination fluctuation and shadow. The sampled rank correlation is a sampled version of the rank correlation combined with median evaluation. The method is also robust for occlusion and ill-conditioned brightness in cluttered scenes. The proposed method is proved to be effective and to have enough precision of location through experiments with images from real world.

View full abstract

This paper will show a new method to embed data into palette-based images while they are generated from the original full-color images. Other embedding methods do not modify the palette of the image. We will show that well-organized palette makes it possible to embed a lot of data into the image with little deterioration in quality. In our method, first, pixels of the original image are divided into n (n <256) subsets using ordinary color reduction method. Then, subsets which have more pixels than others are divided so as to make 256 colors. As the results of experiments, the quality of images made by our method is not inferior to that made by ordinary color reduction without embedding.

View full abstract

This paper proposes a thresholding technique to segment an objective area that has a large difference of gray level and a long boundary between the background area in a gray-scaled image. The average of the gray levels of the pixels with the edge strength that has the maximum frequency in the image is set as the threshold for binarization. The experimental results show that most resultant images by the proposed method were similar to the results by the conventional method that is based on the discriminatory analysis. On the other hand, characters were segmented clearly by the proposed method in the experimental images with shading. The proposed method is expected to apply to various gray-scaled images because it has the simple process and it can determine a threshold automatically without any external parameters to control the process.

View full abstract

This paper describes a method for reconstruction of registered multi-view range images by means of the geodesic dome type of data structure. It enables us to decrease time for storing 3-D shape data into the geodesic dome type of data structure through the 2-D array type of data structure based on sphere coordinate system. The geodesic dome type of data structure ensures the isotropy of sampling area and so it may faithfully represent 3-D shape data. Moreover, it is also remarkably useful as data structure for object recognition, gesture recognition and so on, because it enables us to reconstruct extended gaussian image at the same time with reconstruction of 3-D shape data. Experiments on processing time for reconstruction and accuracy of reconstructed data ensured the validity of the proposed method.

View full abstract

A walk-through technology of virtual environment has been in the spotlight, because the capability for processing three-dimensional CG has been increasing in recent years. In constructing large-scale virtual space, a number of polygons in virtual space cause deterioration of drawing speed. In walk-through of virtual space, a billboard technique has been well known to improve drawing speed maintaining display quality. In this paper, we propose a display method with a cylindrical billboard. In addition, we have implemented a walk-through system of virtual space with proposed method by a client server model and experimental results show that our method is effective.

View full abstract

The detection of motion information is a fundamental research in the field of computer vision or moving picture analysis. In this paper, the precisely detecting method of optical flow is described. First, the optical flow constraint of gradient is calculated and optical flow vector is estimated using the pixel-wise matching procedure. Next, motion of the object is approximated using the parameter of affine-model. The use of these parameters improves the precision of optical flow. The validity of this method is confirmed by computer simulation. It is also examined that this method is further applied to the estimation of distance in depth.

View full abstract