For the motion degraded image of the non-uniform linear motion velocity, there is a method using a power spectrum ratio for the estimation of the motion degradation direction. However this estimated degradation direction have a tendency to be an incorrect result at the peripheral parts of image. This reason comes by the use of the power spectrum ratio between orthogonal directions there, and 0 or 90 degree is erroneously obtained in almost cases. So, in this paper, we propose a method to apply the Hough transform to the multi-level image for solving this problem. When the multi-level Hough transform is used, the power spectrum value obtained in a small evaluation area might contain unsuitable data as a straight-line element depending on the level of the degradation image. In this proposed method, the data being over the most frequent value obtained from the histogram of the power spectrum is adopted to reduce this defect. As a result, the estimation performance for the degradation direction has improved compared with the method by a power spectrum ratio.

View full abstract

This paper presents a 3D foot measurement system for tailor-maid shoes, which using mesh net that is suitable for measurement of a human shape: ‘foot’. The foot is covered with mesh net, and is recorded as images by eight digital still cameras. The coordinates of mesh intersections are extracted by the developed filtering process, and are used here for the matching points of stereophonic measurement. The typical three-dimensional measurement system consists of the laser device and the high-speed image-processing device. These devices are expensive, and general users can' t get the whole of three-dimensional shape easily. We adopted mesh net so that the user could get the whole of three-dimensional shape easily. The user can decide the necessary number of camera and the position easily by referring to the indicated stereophonic area in advance. The measurement points are mesh intersections that can be constructed as the structured data because it is a lattice structure. The structured data is useful for analysis of human foot shape.

View full abstract

This paper describes three-dimensional recovery and analysis method of human motions in a large field employing a mobile frame. Modeling 3-D human motions is requested in various fields such as virtual reality, sport analysis and medical science. Generally, 3-D motion recovery is done by employing stereo vision or factorization based on an extended measurement matrix. These techniques normally employ fixed cameras for taking video image streams. They are therefore difficult in recovering the motions in a large field. In this paper, we propose a technique for achieving 3-D recovery of human motions in a large field employing a mobile set of cameras. Three video cameras mounted on a mobile frame track and capture a person's images from three different directions. The 3-D locations of feature points attached on the object are calculated by factorization with every sample time. Using the proposed technique, several human wide range motions are recovered in a 3-D way. Analysis of the motions shows significance of the technique.

View full abstract

In regards to image stabilization, an electric image-stabilizing system utilizing only video signal processing has achieved small-sized and highly reliable system. However, picture resolution is spoiled during image stabilization in this system. And it is not easy to get a sufficient compensation area when the optical zoom magnification is enlarged. In this paper, a high-resolution electric image-stabilizing system and a combined system which is implemented with the high-resolution electric image-stabilizing system and an optical image-stabilizing system are proposed as methods for improvement of picture resolution and lack of the compensation area. In the former system, adopting a CCD image sensor which has pixels over a conventional CCD image sensor made it possible to get a high picture resolution. Moreover, the latter system has achieved a wide compensation area, a high picture resolution and high and stable image-stabilizing rate by using the optical image-stabilizing system together with the high-resolution electric image-stabilizing system.

View full abstract

In this paper, we propose a viewpoint position measurement system which uses one camera and two mirrors for stereoscopic displays. The two mirrors divide one view area of the camera into two to cover the same observation area from the upper and lower direction respectively. So this system work as a vertical stereo measurement system with particular characteristics that the system can solve the synchronous problem of the stereo measurement and cover the 3/4 of the optic angle of a conventional stereo measurement system with two cameras.

View full abstract

In this paper we propose a precision location and speed detection method with an on-board slit camera and simple road surface marks. An on-board slit camera is looking downward and marks for position detection are put on road surface. The detected mark gives vehicle position in each lane. Position related information can be attached by some simple coding scheme. Vehicle direction corresponds to inclination of the mark in the image and vehicle speed is given by longitudinal length of mark in slit image. Cross mark consisting of four separate bars is selected as a road surface mark for position detection. The position will be given in sub-pixel accuracy as a crossing point of two straight lines calculated by least squares method. Experiments with passenger vehicle and experiment for accurate evaluation show effectiveness of this method.

View full abstract

Embedding of the additional information on the various media can be used for judging if the certificate is forged or not. Authors have already proposed the method of constructing the image information with restoration function on the paper medium. By applying the coding theory, this technique enables an original image information be reproduced from the image information on the paper medium, which got the stain. In this paper, we propose a new technique which can detect the forgery of the digital image data by applying both coding theory and cryptology. We show that it is important to combine error correcting code and cryptogram in an appropriate way. We also present an example of new techniques that use RSA cryptogram.

View full abstract