Conventional retrieval methods of image database (IDB) systems are utilizing non-image information, or alphanumerical data, as keywords. These methods are suitable for document database systems, but not for IDB systems because the contents of image cannot be expressed by only such keywords. In this paper, a new image retrieval method using the layout structure" is proposed. This method can realize image retrieval based on the contents of image. Because, the layout structure means structural information such as object location and spatial relation and it corresponds to composition" of the image. This information, called region feature description (RED), is obtained in the input process of original images. In the retrieval process, we use some specified files which are created form RFD on the basis of each structural feature. Thus, interactive image retrieval based on structural information, that is composition", can be performed. Since India ink paintings are adopted as target images in this paper, it is difficult to extract and describe structural feature of object regions as they are. The structural information is described as geometrical feature by approximating them to simple polygons. As a result, composition" of the target image is obtained. This paper describes the description scheme and the extraction procedure of the structural information. And some experimental results are also presented to confirm the effectiveness of this method.

Analysis of satellite images requires classificatio of image objects. Since different categories may have almost the same brightness or feature in high dimensional remote sensing data, many object categories overlap with each other. How to segment the object categories accurately is still an open question. It is widely recognized that the assumptions required by many classification methods (maximum likelihood estimation, etc.) are suspect for textural features based on image pixel brightness. We propose an image feature based neural network approach for the segmentation of AVHRR images. The learning algoriothm is a modified backpropagation with gain and weight decay, since feedforward networks using the backpropagation algorithm have been generally successful and enjoy wide popularity. Destructive algorithms that adapt the neural architecture during the training have been developed. The classification accuracy of 100% is reached for a validation data set. Classification result is compared with that of Kohonen's LVQ and basic backpropagation algorithm based pixel-by-pixel method. Visual investigation of the result images shows that our method can not only distinguish the categories with similar signatures very well, but also is robustic to noise.

Several types of machines have been proposed to improve the performance of data base management systems, especially the relational one. Among them, a cellular logic type data base machine such as RAP is characterized by its simple structure. Whereas this type of machine outperforms the conventional DBMS software by orders of magnitude for the relatively light load operations such as selection and update, it exhibits poor performance for the heavy load operations such as join and projection. This is because it adopts the nested loop algorithm which is very inefficient. Join has been so far the major performance bottle neck. In this paper we propose a performance enhancement mechanism for the cellular logic data base machine. A novel relational algebra processing algorithm based on the dynamic clustering feature of hash is presented. By introducing the hashing hardware, join operation is much accelerated in comparison with the conventional cellular logic data base machine. Its execution time is evaluated by simulation experiments. It has also been a major problem to handle large relations which cannot fit into the cell memories, where frequent paging degrades the performance severaly. A bucket based staging scheme has been proposed, through which the enhanced architecture can perform the join of large relations efficiently.

The techniques for imaging optically opaque region using an electromagnetic wave radar are being developed. One important application of these techniques is the detection of buried pipes and cables. The image quality of subsurface radar often becomes low because the electromagnetic waves are affected by the attenuation and inhomogeneity of soil. Hence, a method which improves the quality of the radar images has been required. The migration method is utilized in reflective seismic processing and is derived based on the solution of the wave equation represented in spatial frequency domain. It is classified into the F-K and the phase-shift (P-S) migration method. The former is derived on the assumption that propagation velocity of the wave is uniform in the soil while the latter is assumed that the propagation velocity is varying depending on the depth from the ground surface. The P-S method gives relatively good quality images but it requires very long computation time. In this paper, we propose the block migration method in which the F-K method is applied to the divided image blocks with local propagation velocity. In order to solve a problem concerning the connection between the contiguous blocks we present two approaches which are the processings using the overlapped regions and the Lapped Orthogonal Transform (LOT). Some experimental results point out that the block migration method has a good capability of improving the image quality and the processing time using LOT becomes one tenth in comparison with the P-S method.

This paper presents an overview of the SDC-I (Super Database Computer I) developed at the University of Tokyo, Japan. The purpose of the project was to build a high performance SQL server which emphasizes query processing over transaction processing. Recently relational database systems tend to be used for heavy decision support queries, which include many join, aggregation, and order-by operations. At present high-end mainframes are used for these applications requiring several hours in some cases. While the system architecture for high traffic transaction processing systems is well established, that for adhoc query processing has not yet adequately understood. SDC-I proved that a parallel machine could attain significant performance improvements over a coventional sequential machine through the exploitation of the high degree of parallelism present in relational query processing. A unique bucket spreading parallel hash join algorithm is employed in SDC, which makes the system very robust in the presense of data skew and allows SDC to attain almost linear performance scalability. SDC adopts a hybrid parallel architecture, where globally it is a shared nothing architecture, that is, modules are connected through the multistage network, but each module itself is a symmetric multiprocessor system. Although most of the hardware elements use commodity microprocessors for improved performance to cost, only the interconnection network incorporates the special function to support our parallel relational algorithm. Data movement over the memory and the network, rather than computation, is heavy for I/O intensive database processing. A dedicated software system was carefully designed for efficient data movement. The implemented prototype consists of two modules. Its hardware and software organization is described. The performance monitoring tool was developed to visualize the system activities, which showed that SDC-I works very efficiently.