Fumiaki Tajima

Light scattering from a birefringent cylinder, spider silk, slimmer than the wavelength approaches dipole radiation

Using a plane-polarized laser of wavelength lambda = 543.5 or 441.6 nm and spider silk, we investigated the diffraction of a transparent cylinder of diameter D approximately lambda - lambda/5 at normal incidence. The measured pattern corresponded well to the one calculated by a rigorous solution of the theory for the problem. The birefringent index and D of the sample could be determined simultaneously. The experimental data of the scattering cross section for D < lambda/4 suggested that the data approached dipole radiation.

Low-pass Filter’s Effects on Image Analysis Using Subspace Classifier

This paper shows an effect for applying a low-pass filter on the performance of image analysis using the Subspace classifier. The feature extraction was firstly based on three kinds of intensity distributions, and the feature vector and subspace dimension for recognition were examined. Afterwards, a series of the analysis on the accuracies were conducted in the cases of filtered images and without filtered. The analyzed accuracies by using the Subspace classifier were also compared with the results by the technique of another: Learning vector quantization (LVQ).

Standard optical coaxial double fiber diameter and refractive index measurement, accuracy, and precision using light scattering at normal incidence

A plane-polarized laser wave with a wavelength of 441.6 nm illuminates a cladding optical fiber with a diameter of about 18 to 38 mum at normal incidence. A measured angular distribution of the intensity of the scattered wave corresponds well with the differential cross section of a rigorous theoretical calculation of a coaxial double cylinder over a wide range of scattering angle. The diameter and refractive index of the cladding and core of the illuminated part of a fiber have been determined accurately for each uncertainty.

Filtering Effects for Image Data Types in Image Analysis Using Subspace Classifier

This paper describes an effect on performance of classification with applying of a low-pass filter to an image of a few of distinct image types in the image analysis using Subspace classifier method. The feature extraction was firstly examined based on three kinds of intensity images, and for classification, the feature vector and Subspace dimension were determined. Afterwards, the images with a few of distinct image types were analyzed for classification performance, and filtered images were also analyzed. The analyzed accuracies of filtered images were compared with the accuracy without filtering. Our results showed that the features of true-color channel were suitable for classification, and that an application of filter to an image of a few of distinct image types had an influence on a classification accuracy.

Composition of Learning Patterns Using Spherical Self-Organizing Maps in Image Analysis with Subspace Classifier

This paper describes a composition of learning patterns based on the visualization of a Spherical SOM for improving the performance of image analysis using the Subspace classifier. We have applied the Subspace classifier to image analysis because it has fewer parameters and higher performance. Then we have experienced that the selection of features and learning patterns influence greatly the classification performance through examinations. The Spherical SOM has no border in the array of nodes and eliminates the Border effect problem. Comparing the performance of the image analysis, we show that visualization of the Spherical SOM allows the composition of learning patterns to improve more performance and degree of its reliability than those without the composition.

Classification of Fundus Images for Diagnosing Glaucoma by Self-Organizing Map and Learning Vector Quantization

This paper presents a two stage diagnosis system that consists of Self-Organizing Map (SOM) and Learning Vector Quantization (LVQ) subsystems for diagnosis of fundus images. The first stage performs clustering and pseudo-classification of the input feature data by a SOM. The use of the pseudo-classes is able to improve the performance of the second stage consisting of a LVQ codebook. The proposed system has been tested on real medical treatment image data. In the experiments we have achieved a maximum accuracy rate of 71.2%, which is comparable to other results in literature.

Image Analysis of Fundus Using Filter Bank and Subspace Classifier

This paper describes filtering effects on classification performance with applying of filter bank of multi scale and orientation to an image of distinct types in the image analysis using Subspace classifier method. In our proposed method, the feature extraction was firstly conducted, and examinations on the feature vector and subspace dimensions were conducted based on three kinds of intensity distributions. Afterwards, a series of the analysis concerning the accuracy were conducted in the cases of single filter and filter bank. The effects on the performances for filter bank were compared with the performances for a single filter. Our results showed that the preprocessing using filter bank provided a higher classification performance than the results using a single filter preprocessing, and that boosted the improvement of classification performance.

Scattering of a light wave by a thin fiber on or near a prism: experiment and analytical theory

We have performed an experiment of the scattering of the near field on a prism created by a laser wave, evanescent wave (EW), or plane wave (PW) of an incident angle slightly larger than or smaller than the critical angle, by a thin fiber of subwavelength diameter set above the prism, and we made an analytical theory of an adapted model for the experiment. We have been able to analyze the experimental data exactly by the model theory better than any other theory we have ever known. The importance of the multiple interaction of the wave between the fiber and the surface and also the close similarity of the scattering characteristics between the EW and the PW mentioned above have been acknowledged by the analysis of the data obtained.

Applying the significance degree by SOM to image analysis of fundus using the filter bank

This paper describes the filtering effects on classification performance with applying significance degree by SOM to the image analysis using filter bank preprocessing and Subspace Classifier. In our proposed method, a series of analysis concerning accuracy were first conducted in the cases of single filter and filter bank, and examinations on significance degree by SOM were conducted based on green(G) and blue(B) color channels. The difference of the filtering effect between two color channels was compared with using the significance degree. We show that the difference of the filtering effect between two channels can be clarified by using the significance degree by SOM.

Real part of the dielectric constant of a subwavelength-in-diameter silver pipe is positive in visible light

The complex refractive index of a silver pipe of which the diameter is subwavelength in diameter have been found to be different from those of a silver thin film for the first time. The silver pipe is made of cladding a spider silk of a half-wavelength diameter with silver. It is illuminated by a plane-polarized laser of wavelength 660 nm at normal incidence. The angular distribution of the light intensity scattered by the pipe is measured and fitted using theoretical calculations based on the corresponding model. The fitting results have led to the optimum values and uncertainty ranges of the index and the diameter of the pipe.