Kazuo Toraichi

Feature extraction of handwritten Japanese characters by spine functions for relaxation matching

Abstract This paper proposes a method of extracting features of handwritten Japanese characters by using spline approximation. This method represents a character by contours expressed by well-approximating functions and stable breakpoints which characterize the connection of the strokes so that it provides proper features for handwritten Japanese character recognition with relaxation matching. Its effectiveness tested on a Japanese character database is also shown.

On a grading system for beef marbling

Abstract This paper presents an automatic grading system for beef marbling which is composed of discriminant threshold selection method and run length processing. From the experimental results, it has been confirmed that the proposed system enables high quality grading of beef marbling, and robust region segmentation of the actual beef rib-eye image into lean and fat regions.

Two-dimensional spline interpolation for image reconstruction

A scheme of two-dimensional (2-D) spline interpolation for image reconstruction is presented. It is aimed to be applied in image data compression. An impulse response derived from 2-D quadratic spline functions is theoretically analysed. Then it is approximated to be a finite-duration one (FIR) for practical implementation. With this scheme, compression rate of greater than 12 is achieved for X-ray left ventricular cineangiogram image data compression.

A smooth signal generator based on quadratic B-spline functions

A hardware system of quadratic spline interpolation is proposed. Its properties are that switching noise with digital-to-analog (D-A) converter chips does not appear and that the output is smooth without any analog lowpass filters that disorganize the phase characteristics. >

On speeding candidate selection in handprinted Chinese character recognition

Abstract A method of selecting candidate categories to speed handprinted Chinese character recognition is proposed. This method is set up in the pre-stage of conventional candidate selection and reduces the large candidate selection time which is due to a vast number of Chinese categories. Its performance tested on a Chinese character database is also shown.

On a fluency image coding system for beef marbling evaluation

This paper presents a Fluency Image Coding System of beef rib-eye images for Beef Marbling Evaluation. This is the second in a series of cooperative researches with the Japan Livestock Technology Association under an initiative to construct an Automated Online Beef Marbling Grading Support System by image analysis techniques. Our first cooperative research was on a Beef Marbling Grading Method, and was published in this journal in [Pattern Recognition Lett. 21 (12) (2000) 1037-1050]. This second cooperative research focuses on a binary image coding system that supports remote observation of beef marbling structure from a database of coded beef rib-eye images by users including meat graders, livestock producers, and researchers. Image encoding is by a novel automatic contour compression method based on function approximation via interpolation using the Fluency Compactly Supported Sampling Functions of degree 2. Image decoding, based on interpolation of the encoded data by the similar functions, enables the web-browser based decoder to reconstruct the original fat contours smoothly even on Affine-transformed enlargement. Experimental results showing, respectively, size and image quality comparisons with other formats that support binary images and several enlargement schemes are included for evaluation.

A parallel architecture for relaxation operations

Abstract Relaxation method attracts attention as an effective parallel method in image processing and pattern recognition. But their previous parallel implementations are specialized to each application such as image processing at pixel level. The present paper proposes a parallel architecture for a relaxation processor which can be applied to general use. This architecture is based on the round-robin structure of data communication among the processing elements. Its effectiveness is discussed by evaluating its performance when applied to a pattern recognition method.

Automatic seal identification using fluency function approximation and relaxation matching method

An automatic seal identification method using a relative relation among segments of the seal impression is proposed. Features of a seal impression image are extracted by fluency function approximation, and matching is performed by relaxation matching. 2 (seals) /spl times/ 2 (sets) images are used to verify the performance of the present method. The recognition rate is 88.8%, and the processing speed for observing and matching is about 30 s on middle scale computers.<<ETX>>

On a relaxation-labeling algorithm for real-time contour-based image similarity retrieval

In this paper, we propose a relaxation-labeling algorithm for real-time contour-based image similarity retrieval that treats the matching between two images as a consistent labeling problem. To satisfy real-time response, our algorithm works by reducing the size of the labeling problem, thus decreasing the processing required. This is accomplished by adding compatibility constraints on contour segments between the images to reduce the size of the relational network and the order of the compatibility coefficient matrix. Particularly, a relatively strong type constraint based on approximating contour segments by straight line, arc, and smooth curve is introduced. A distance metric, defined using the negative of an objective function maximized by the relaxation labeling processes, is used in computing the similarity ranking. Experiments are conducted on 700 trademark images from the Japan Patent Office for evaluation. q 2003 Elsevier Science B.V. All rights reserved.

Window functions represented by B-spline functions

Characteristics of window functions provided by B-spline functions are analyzed in comparison to other known windows. For the same highest-sidelobe level, those windows have a mainlobe bandwidth between those of Gaussian and cos/sup a/ x windows are as good as those of cos/sup a/ x ones in the sidelobe falloff. >