Toshiyuki Sakai

Color information for region segmentation

Abstract In color image processing various kinds of color features can be calculated from the tristimuli R, G, and B. We attempt to derive a set of effective color features by systematic experiments of region segmentation. An Ohlander-type segmentation algorithm by recursive thresholding is employed as a tool for the experiment. At each step of segmenting a region, new color features are calculated for the pixels in that region by the Karhunen Loeve transformation of R, G, and B data. By analyzing more than 100 color features which are thus obtained during segmenting eight kinds of color pictures, we have found that a set of color features, (R + G + B) 3 , R − B , and (2G − R − B) 2 , are effective. These three features are significant in this order and in many cases a good segmentation can be achieved by using only the first two. The effectiveness of our color feature set is discussed by a comparative study with various other sets of color features which are commonly used in image analysis. The comparison is performed in terms of both the quality of segmentation results and the calculation involved in transforming data of R, G, and B to other forms.

Line extraction and pattern detection in a photograph

Abstract An experiment on line extraction and pattern detection in a photograph is performed. From digitized pictures which have eight gray levels, the value and direction of the steepest descent is first calculated for each point with 3 × 3 meshes. Then those points whose gradient is more than certain threshold value are regarded as component points of line segments. From these points the estimation is done to extend the line portions as far as possible to the neighbouring ones. The pattern detection is done mainly for the photographs of human faces which are in front direction. A standard pattern of a front face is stored in a computer and the correlation values with several sizes of the standard pattern are calculated for the face contour, eyes, nose, and mouth independently. From these correlation values of the face portions, the final determination is done for the existence of a face. This two stage recognition process is adopted because certain portions of a face may be enough for the decision even if the other portion of a face is hidden in noise and could not be detected as such. For many photographs of human faces in several conditions the whole algorithm has been tested with considerable success.

Efficient diagram understanding with characteristic pattern detection

Abstract A connection diagram understanding system using a facsimile as its input device has been designed and implemented. The principle of this system is described and examples of its application to some hand-sketched diagrams are shown. In order to reduce processing time, the procedure is started by only accessing the pixels located on the borders of certain meshes to detect characteristic patterns and make a control map. Background judgment and long straight line segment extraction are executed on the control map. Other complicated areas, which are usually a small portion of the whole diagram area, are also indicated by special labels on the map and then processed by a detailed procedure which scans every pixel at these areas. Graph descriptions of diagrams are employed at different steps of the hierarchical understanding. Problems of data compression, diagram retrieval, and diagram editing are discussed.

The Automatic Speech Recognition System for Conversational Sound

This paper describes the method and the system investigated to solve the problem encountered in the automatic recognition of speech sound. From research in the automatic analyzer of speech sound, a monosyllable recognition system was constructed in which the phoneme is used as the basic recognition unit. Recently this system has been developed to accept the conversational speech sound with unlimited vocabulary. The mechanical recognition of conversational speech sound requires two basic operations. One is the segmentation of the continuous speech sound into several discrete intervals (or segments), each of which may be thought to correspond to a phoneme, and the other is the pattern recognition of such segments. For segmentation, by defining two criteria, ``stability and ``distance, the properties of the time pattern obtained by the analysis of input speech sound may be examined. The principle of the recognition is based on the mechanism of the articulation in our speech organ. Corresponding to this, the machine has the functions called phoneme classification, vowel analysis and consonant analysis. A conversational speech recognition system with the phonetic contextual approach is also applied to the vowel recognition where the time pattern of input speech is matched with the stored standard patterns in which the phonetic contextual effects are taken into consideration. The time pattern which has great variety may be effectively expressed by the new representation of ``sequential pattern and ``weighting pattern.

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.

Extraction of invariant picture sub-structures by computer

This paper treats the problem of determining the structure of a picture—in particular, detecting the same structure in two different pictures. The content of the pictures is restricted so that it is not necessary to treat curved lines. Line elements are first extracted that reflect enough information about the original pictures; they are represented by a list structure that can be handled by the LISP programming language. Next, the presence of the same topological substructures in the two pictures is detected. Finally, coefficients defining the transformation between these structures are determined. This procedure has been tested successfully on a wide range of pictures having simple structures.

Fundamental algorithm for train scheduling based on artificial intelligence

The trial-and-error efforts by experts have been considered as indispensable in compiling railroad timetables. There is a strong requirement to computerize the procedure, but it has been considered difficult. It is proposed to merge the following two concepts as a means to obtain an efficient trial-and-error procedure, which is the basis of the compilation of railroad timetables: (1) the concept of (time) belt, where a number of alternatives for arrival and departure is handled as one alternative, considering only the order of arrivals and departures; (2) the concept of the propagation of constraints, which is well known in the artificial intelligence where alternatives in conflict to the order of arrivals and departures are eliminated. This paper discusses the scheduling problem for the partially ordered set of events. The requirement for the time difference between events is considered as a constraint. An algorithm is proposed which calculates the permissible range of time for the event generation, as a closed interval on the time axis. It is shown that the constructed algorithm can be applied as the basic algorithm for the railroad timetable compilation system, where the timetable is considered as a specification for the order of arrivals and departures of the trains, and the arrival-departure time is considered as a belt. An efficient method of compilating the timetable by the proposed algorithm is presented, based on the search for the possible orders of the trains, together with the construction example of the timetable for the actual railway.

Automatic model generation and recognition of simple three-dimensional bodies

A method of recognizing polyhedra is described which matches “noisy models” to the models derived from analyzing the real scene.

Acoustic noise reduction by two dimensional spectral smoothing and spectral amplitude transformation

We will describe a noise reduction method which diffuses and suppresses the noise component, while concentrating and enhancing speech formant information. Two Dimensional Spectral Smoothing (TDSS) is applied over a time sequence of spectral envelope to diffuse noise and concentrate formant information. After consonant enhancement, Non-linear Spectral Amplitude Transformation (NSAT) is carried out to further suppress the noise component and enhance formant information. This noise reduction method has proven to be superior to the conventional frequency subtraction method.

Simulation of traffic flows in a network

A computer simulation program which deals with traffic flows in the network of a large area is described. Each road is segmented into blocks of several ten-meter lengths and is represented by a bidirectional list in computer memory. The movement of cars, i.e. the transfer of cars from one block to the next, is expressed by a proper formula. This formula is based on the supposition that the speed of cars in a block is determined only by the density of cars in the block, and this speed-versus-density curve is empirically given the numerical values.nThis simulation scheme has its excellent point in that it makes it possible to trace the dynamic behavior of traffic flows in a variety of situations, some examples of which are given for an actual area of the city of Kyoto, Japan.