The purpose of this study is to make moving images not from illumination-controlled intermittent motion but from physically contiguous motion. If the apparent motion can be produced by step motion and explained by a physical model, a distinction between real motion and apparent motion, which have been thought to be completely different, can be discussed. In our research, we have tried to design new moving image media using a mechanical method as a substitute for the optically controlled method (i.e. slits or strobe). Additionally, experiments were implemented about the perception of the apparent motion produced by step motion with our system. As a result, it became clear that the perception of apparent motion differs between individuals. This theory of moving image perception makes it possible to make moving images by using methods different from the existing one.

抄録全体を表示

If terrestrial digital broadcasting services are available at ships used for transportation, they will be very useful and convenient for gathering information including during disasters. We studied and investigated to realize stable reception of digital TV waves and re-transmission technology at ships connecting Hiroshima-Matsuyama. As a result, BER degradation of unexpected multi-path waves with over guard interval length, adjacent channel interference, and large level changes of electric fields that are characteristics of mobile ship reception occurred. Therefore, shielding undesired waves using the ship structure and selection of receiving antennas in accordance with electric field strength and so on are taken. Through these countermeasures, digital TV waves can be received at 90% on all lines and one segment reception can be available for the whole cabin of a ship by retransmission.

抄録全体を表示

In video abstraction, viewer's emotional reactions to original video play an important role because people often choose what to watch based on the emotional context. We have devised a method for incorporating viewer's perceived impressions in video abstraction. The impression level is measured by the degree of arousal, based on two measures of heart activity: deceleration of heart rate and activation of the higher frequency component of the heart rate variability. For an evaluation of our method, we performed two sets of experiments comparing the abstracts generated by our method against subjective selections. Three short videos with different arousal levels (two arousing and one calm) were used. In Experiment 1, shot matching between two abstracts was examined: the arousing videos showed a 50% match, while the calm one showed a 30% match. In Experiment 2, the abstracts were displayed to the subjects, which they evaluated subjectively: our method generated good results in the arousing/pleasant video.

抄録全体を表示

Recently, programs that use 3D models have become remarkable due to the digitalization of various media, improvement in the performance of computers, and the development of the Internet. Present production technology contains much advanced software that cannot be easily utilized because it takes a considerable amount of time for a general user to master it. In this research, a system was developed that can easily simulate the movements of a finger by using an “action unit” that models as the muscle. The model of the finger constructed using this system is subjectively evaluated.

抄録全体を表示

We attempted to find common expression changes for people with addictive tendencies when they come into contact with addictive substances. We discovered a specific index from a person's change in expression in line with the addictive tendencies and suggest a technique to calculate the degree of addiction. This paper investigates the relationship between facial expressions and dependency. In earlier studies, we found a tendency among addicted subjects to experience changes in desire when coming into contact with something associated with the addictive object. At the same time, their expressions changed as their desired changed. Therefore, we applied a dependency diagnosis support method to these changes in expression. As a result, dependency was able to be detected with an accuracy of more than 75% using the suggested experimental evaluation method. It was shown that the proposed method was effective at diagnosing dependency.

抄録全体を表示

Estimating the noise level of a digital image is important for many applications such as noise removal, image quality assessment, and so on. For practical use in these applications, the estimation method should be achieved from a single image. Ce et al. developed a single image noise estimation method. In their method, a priori information of noise characteristics was generated by the computer simulation and used in Bayesian MAP. However, they did not consider the inherent characteristics of each camera, depressing the estimation accuracy. Therefore, in this paper, we take into account the inherent characteristics of a camera by using camera specific a priori in Bayesian MAP. Camera specific a priori is constructed from preliminarily measurements of the camera instead of a computer simulation. The experimental results show the improved accuracy of noise estimation.

抄録全体を表示

Recently, an efficient video coding method at low bit rate using Matching Pursuits (MP) has been proposed. The MP coding method represents a signal in an approximate form using a dictionary. Therefore, coding performance depends greatly on the dictionary. In this paper, we introduce a video coding method that employs motion compensation and MP using a dynamic learning dictionary. The dictionary of the proposed method is renewed at each frame by using encoded information. Simulation results show that the coding performance of MP can be improved by applying the dynamic learning dictionary.

抄録全体を表示

Vertical orientation on the ventriloquism effect inherent to flat panel display televisions has been evaluated. The stimuli were an image of a metronome and its accompanying sound. These evaluations were aimed at understanding the extent of the effect during real use. The results show that with loudspeakers placed in a low position the effect was too minor to lift a sound image to the upper part on the screen.

抄録全体を表示

This paper describes a method for expressing two-dimensional (2D) discrete data using a series. The coefficients of the series are determined by using the components obtained by taking a 2D fast Fourier transform of the data. The use of the series enabled the following operation to be performed. First, reducing the number terms in the series was equivalent to performing a filter operation on the data. Next, a differential or integral representation equation was obtained by applying a derivative or integral operator to the series. Furthermore, the data was able to be interpolated while preserving its spatial frequency characteristics.

抄録全体を表示