/ Shunji Mori / Professor
/ Jintae Lee / Associate Professor
/ Tony Y. T. Chan / Assistant Professor
/ Yu Nakajima / Research Associate
This year we were active in both research and education. That is, 4 journal papers were published and 4 conference papers were presented. Concerning the education 10 graduation research theses and two master theses were supervised. On the other hand, we contributed domestic companies very much.
Prof. S. Mori conducted big project of Automatic Signature Verification System cooperated with Nihon University, Tohoku University, and Axiom Ltd. , which was supported by NEDO. He instructed Aiko Engineering Ltd. to develop high performance OCR which is specific to medical terms and low quality printed characters, which was sponsored by Tohoku Computer Ltd. He also did for the company to develop an automatic reading system of a teaching evaluation sheet and so contribute to mechanization of the administrative business in Student Affairs division of the University.
Refereed Journal Papers
An adjusted mass-spring model for animating cloth objects, improved in order to avoid unrealistic elongation and to take into account the elongation curve of woven fabrics, is described.A cloth is approximated to a deformable surface composed of a network of masses and springs, and the movement is simluated using the fundamental law of dynamcs. When a concentrated high stress is given to a local region of the surface, unrealistic elongation occurs.With a simple mass-spring model, the only solution is to increase the stiffness of the springs.However, it increases the cost of computation and deteriorates the natural elastic movement at other areas.In this paper, we present a method to adjust the elatsticity of springs to adapt the model to the natural stiffness of textiles.
A technique is presented for synthesis of diffused images based on special paper model used in oriental black ink painting. With the paper model, users can represent wide range of randomly generated fiber meshes, providing a basis for dynamic ink diffusion. The image diffusion algorithm is effective for creating diffusion images produced on different types of paper. The resulting diffusion images reflect the texture and the global change of gray tone as well as the local temporal variations specific to the paper.Examples that demonstrates the capability of the proposed technique are presented for both the liquid ink and paper types.
Almost all methods of chromosome recognition assume supervised training; i.e., we are given correctly classified chromosomes to start the training phase. Noise, if any, is confined only in the representation of the chromosomes and not in the classification of the chromosomes. During the recognition phase, the problem is simply to calculate the string edit distance of the unknowns to the representatives chosen from the training phase and classify the unknowns accordingly. Here, a general method to tackle the difficult unsupervised induction problem is described. A success of the method is demonstrated by showing how the inductive agent learns weights in a dynamic manner that allows it to distinguish between noisy median and telocentric chromosomes without knowing their proper labels. The process of learning is characterized as the process of finding the right distance function, i.e., the distance function that can nicely separate the classes.
This research is an attempt to quantify aesthetical features of Chinese character (Kanji) with the view of finding out some common factors of beautiful form. Assuming that printed characters can be generally accepted as `beautiful' characters, the features of printed Kanji are calculated. The feature indices used here are --- center of gravity, spread, and complexity based on the method of moment. From the numerical features, it was found that `balancing' is one of the important factors of beauty of characters. Also it was found that the balancing process is common to different font designers though the fonts themselves show different appearances.
Refereed Proceeding Papers
A technique is presented for synthesis of diffused images based on special paper model used in oriental black ink painting. With the paper model, users can represent wide range of randomly generated fiber meshes, providing a basis for dynamic ink diffusion. The image diffusion algorithm is effective for creating diffusion images produced on different types of paper. The resulting diffusion images reflect the texture and the global change of gray tone as well as the local temporal variations specific to the paper. Examples that demonstrates the capability of the proposed technique are presented for both the liquid ink and paper types.
A method is proposed for supervised learning to classify bit strings for three classes. The learner was modeled by two formal concepts: transformation system and stability optimization. Even though a small set of short examples were used in the training stage, all bit strings of any length were classified correctly in the online recognition stage. The learner successfully learned to devise a way by means of metric calculations to classify bit strings according to 3-parity-ness while the learner was never told the concept of 3-parity-ness.
A method is proposed for supervised and unsupervised learning to classify bit strings for three classes. The learner was modeled by two formal concepts, transformation system and stability optimization. Even though a small set of short examples were used in the training stage, all bit strings of any length were classified correctly in the online recognition stage. Neural networks, so far, are unable to achieve this result because they assume a finite dimensional vector space. Typically, the complexity of the neural network increases as the dimensionality of the problem increases. The proposed learner, however, successfully learned to devise a way by means of metric calculations to classify bit strings of any length according to 3-parity-ness. The system was able to acquire the idea of counting, dividing, and taking the remainder by evolving a set of string-editing rules along with their appropriate weights.
To find a global and smooth curve fitting, cubic B-Spline method and gathering-line methods are investigated. When segmenting and recognizing a contour curve of character shape, some global method is required. If we want to connect contour curves around a singular point like crossing points, merging separated contours together which lies apart crossing the singular point is necessary. For this purpose, cubic B-Spline method and new line-gathering method are investigated and proposed. The result is that cubic B-Spline method is rather too easy to bend. Easy to bend feature can cover singular point smoothly, so is not good to detect singular points. Gathering-line method is to represent the contour by overlapped line segments. By overlapping lines, arcs can be represented in natural way. Some investigations and experimental results are shown.
Academic Activities