Toshikazu Kato

A sketch retrieval method for full color image database-query by visual example

Gives a basic idea and its fundamental algorithms of the visual interface for image database systems. The QVE (Query by Visual Example) accepts a sketch roughly drawn by a user to retrieve the original image and the similar images. The system evaluates the similarity between the rough sketch, i.e. a visual example, and each of the image data in the database automatically. The QVE interface is implemented and examined on an experimental electronic art gallery called ART MUSEUM. This paper also gives some experimental results and a current evaluation. The algorithms are quite effective for content based image retrieval.<<ETX>>

A similarity retrieval of 3D polygonal models using rotation invariant shape descriptors

In virtual reality and multimedia applications, 3D polygonal models are increasing in number. Similarity retrieval is an important task in 3D polygonal model databases. We present rotation invariant shape descriptors for similarity retrieval. Our feature descriptor grouping technique overcomes the efficiency problem of query processing in high-dimensional shape descriptor spaces. Although high-dimensional feature descriptors are reduced by our techniques, they maintain high recall and precision.

Learning of personal visual impression for image database systems

Visual impression may differ with each person. User-friendly interfaces for image database systems require special retrieval methods which can adapt to the visual impression of each user. Algorithms for learning personal visual impressions of visual objects are described. The algorithms are based on multivariate data analysis methods. These algorithms provide a model on visual perception process of each user from a small set of training examples. This model is referred to as a personal index to retrieve desired images for the user. These algorithms were implemented and examined in a graphical symbol database system called TRADEMARK and a full color image database called ART MUSEUM.<<ETX>>

MACSYM: A hierarchical parallel image processing system for event-driven pattern understanding of documents

Abstract MACSYM is a hierarchical parallel processing system for pattern understanding applications. It features event-driven parallel processing for knowledge-based understanding of document images. The system is composed of a master processor, slave processors and a large shared memory, and is equipped with versatile communication facilities. The parallel processing software system M.UM has been developed on MACSYM. It supplies a parallel processing language MacC, an extended version of C, and supports the programming for document image understanding. The Japanese newspaper layout understanding system EXPRESS is being developed on MACSYM. It analyzes a newspaper image and extracts articles in a few seconds.

Image retrieval system using impression words

This paper introduces the image database system which can retrieve using impression words. ART MUSEUM (Multimedia Database with Sense of Color and Construction upon the Matter of ART) is our prototype system for advanced image database. It allows users to retrieve images from a database of impressionist paintings by using adjective words. The system learns the correlation between the feature of sample paintings and users descriptions, and provides interactive and progressive learning mechanisms. We show the interesting experimental results on ART MUSEUM system.

Visual Interaction with Electronic Art Gallery

This paper describes the ideas of visual interaction and their implementations on an image database system. The visual interaction includes a query by visual example (QVE) and a query by subjective descriptions (QBD). The former provides a sketch retrieval function and the latter provides a sense retrieval function. This paper shows the detail algorithms of these functions developed on our experimental database system ART MUSEUM. These functions use a pictorial index created by image analysis and a personal index automatically learned as the user model. They formed visual interaction in a truly user-friendly manner. This paper also summarizes the requirements to a multimedia data model from a cognitive aspect.

A high fidelity contrast improving model based on human vision mechanisms

We propose a computational model of adjusting color contrast and luminance contrast nonlinearly, dynamically and adaptively based on the image content. It uses lateral inhibition mechanisms and adaptation (light/dark, color) mechanisms. Our model can improve the appearance of an image naturally, both locally and globally at the same time. Our transformation shows a reverse-S-shape response, with the curvature being determined based on the content characteristics of each individual image. We demonstrate the performance by image enhancement experiments. It shows good performance in not only luminance, but also chromaticity. Our model and transformation can be applied to many fields, especially, storing or repairing images of art or cultural heritage, because it can improve images preserving original subjective impression.

3D object retrieval based on subjective measures

To support the creation of good virtual world presentation, we need to retrieve polygon objects from databases. In conventional database systems, retrieval of polygonal data was based on only the features of the polygon. Unfortunately, such retrieval techniques do not consider the users subjective measures. This paper illustrates the creation of multidimensional spaces based on subjective measures by analyzing the correlation between features of polygonal data using a multidimensional scaling method and regression analysis. Our techniques enable users to retrieve objects more efficiently by considering subjective measures.

Multimedia interaction with image database systems

A picture is worth a thousand words. Images are good media to communicate with an image database system. Multimedia interaction enables users to give multimedia information to the system as well as to receive it from the system[1, 2]. Visual interface must be able to interpret several kinds of multimedia information related with computer and human vision. For example, the system should recognize the words on subjective impression as well as the sketches we draw as the keys for retrieval.

Modeling of subjective interpretation for street landscape image

Our hypothesis is that the model of subjective interpretation is based on rough information process. We analyze a relationship between adjectives and three graphical attributes of color, direction, and position in order to model the subjective interpretation for streetscape images. We focus especially on the graphical descriptors such as Lab color space, Lab auto-correlation and rough image segmentation, which correspond to the graphical attributes. Our psychological experiment indicates that subjective interpretations consist partly of “evaluation”, “texture”, and “openness”. By analyzing the relationships between these three factors and the graphical descriptors, we find that subjective interpretation relates to the combination of the three graphical descriptors.