Yasuhiro Ohmiya

Image Resolution Conversion Based on an Edge-Adaptive Interpolation Kernel

In this paper we propose a novel edge-adaptive interpolation method which is easy to implement and computationally efficient. Although various edge-adaptive image interpolation methods have been proposed in the past, none of them are comparable to classical interpolation methods such as Bi-linear interpolation or Bi-cubic interpolation concerning computational cost. Here we propose a novel interpolation method modifying the classical interpolation methods adaptive to improve the quality around the edges. The proposed methods flow is the following: 1) For all pixels, determine the local gradient defined by plane approximation of four neighbouring pixels. 2) Derive the edge-adaptive interpolation kernel by edge-directed interpolation axes. 3) Interpolate. From the results of experiments, the effectiveness of the proposed method concerning the computational cost and subjective image quality have been verified.

Proposal of contents-oriented cognitive network architecture towards realization of the New Generation Network

This paper proposes a contents-oriented cognitive network architecture as post-Internet. This architecture consists of three layers, that is, physical, semantic and application layers. Fluency function approximation-based unified coding of multimedia is one of key technologies. In semantic layer, both of cognitive function which corresponds to “intuition function”, that is, “pattern recognition” and logical processing are implemented. A contents-overlay, which consists of same kinds of contents based on the criterion of contents-similarity, can be formed. The unified coding of multimedia is effective for not only explosion of information but energy saving. Moreover, contents-overlay is effective for SPAM and for lightening the burden on network administrators

A Function Approximation Method for Images with Grading Regions

In the last decade, many function-approximation methods for illustration images have been proposed. However, these methods cannot process more complex full color images effectively, even when some simple regularity of color transitions such as gradations appear in images. In this paper, we propose a rigorous function-approximation method that is applicable for gradation images. This method performs image segmentation that recognizes a gradation pattern as one region. This segmentation is performed by recursively approximating the tones of colors using multiple regression analysis of 2-variable functions. We compare the proposed method and the conventional methods with regard to the quality of enlarged images and the efficiency of image description and reproduction. The experimental results show that our approach is higher in quality, smaller in data size, and faster in computational costs than the conventional methods.

Working ratio estimation method for a long distance fixed wireless access (22 GHz band, 6.3 Mbit/s) link

This paper discusses a method to determine the optimal node location before a long distance FWA network (22 GHz band, 6.3 Mbit/s) is set. During our 3 years experiment, we found out that the long distance FWA net with 2 links (about 60 km) was down for approximately 42 hours per year due to heavy rain fall conditions. This means that the long distance FWA net is practicable to some extent. We discuss a method to estimate the working ratio between two unknown locations using AMeDAS weather data. By using this method, an appropriate node location can be decided before FWA net is deployed. This approach can improve the working ratio in the sense of link-up continuation.