This paper describes technological and operational issues of an image-art-on-demand system, which provides visitors with high-definition images of fine art in a virtual gallery. The system is presented as a typical example of multimedia information service systems on IP over ATM network. The high-definition images of fine arts from a database are interactively selected in a virtual gallery which is generated by an advanced computer graphics (CG) workstation. The generated images of the virtual gallery are transmitted by MPEG-2 over TCP/IP on ATM at 30 frames per second. This system was opened from January 1996 to March 1997 as one project of NTT's joint utilization tests of multimedia communications. As far as we know, this system is the first real-time image-art-on-demand system using MPEG-2 on IP over ATM-WAN to be exhibited to the general public.

We present a volume rendering algorithm which renders images at approximately two to seven times the speed of a conventional ray caster with almost no visible loss of image quality. This algorithm traverses the volume data in object order and renders the image by performing ray casting for the pixels within the footprint of the voxel (i.e., rectangular prism) being processed. The proposed algorithm supports the rendering of both single and multiple isosurfaces with arbitrary opacity values. While the projection approach to volume rendering is not new, we present an algorithm specifically designed for the perspective projection, evaluate its rendering speed for both single and multiple isosurfaces with arbitrary opacity values, and examine how efficiently it uses cache memory.

When people watch such motion pictures as documentaries or educational-type films, it is very natural for them to be interested in moving objects in the movies and be eager to know the detailed information related to these object. Therefore, a mechanism that enables users to directly pick up object information from motion pictures is necessary to make a movie system feasible. For this reason, we are researching techniques on using objects in motion pictures as hypermedia anchors. We call a movie system that provides the above mechanism a video hypermedia system. An object in a motion picture can generally be considered as a complex object which includes many parts. To allow users to obtain information related to each part, a system must be able to provide anchors corresponding to each part in each complex object. For this, authors cannot help defining all anchors in all frames, since the visual status of each part varies from moment to moment. This paper presents our approach for managing objects in motion pictures for video hypermedia systems. The main feature of the proposed method is to apply computer graphic techniques to the defining of anchors for a complex object.

In this paper we propose a new medical image processing system called Virtualized Endoscope System (VES)", which can examine the inside of a virtualized human body. The virtualized human body is a 3-D digital image which is taken by such as X-ray CT scanner or MRI scanner. VES consists of three modules; (1) imaging, (2) segmentation and reconstruction and (3) interactive operation. The interactive operation module has following thee major functions; (a) display of, (b) measurement from, and (c) manipulation to the virtualized human body. The user of the system can observe freely both the inside and the outside of a target organ from any point and any direction freely, and can perform necessary measurement interactively concerning angle and length at any time during observation. VES enables to observe repeatedly an area where the real endoscope can not enter without pain from any direction which the real endoscope can not. We applied this system to real 3-D X-ray CT images and obtained good result.

Various kinds of new architectures have been proposed to enhance operating performance of the DRAM. This paper reviews these architectures including EDO, SDRAM, RDRAM, EDRAM, and CDRAM. The EDO slightly modifies the output control of the conventional DRAM architecture. Other innovative architectures try to enhance the performance by taking advantage of DRAM's internal multiple bits architecture with internal pipeline, parallel-serial conversion, or static buffers/on-chip cache. A quantitative analysis based on an assumption of wait cycles was made to compare PC system performance with some architectures. The calculation indicated the effectiveness of external or on-chip cache. Future trends cover high-speed I/O interface, unified memory architecture, and system integrated memory. The interface includes limited I/O swing such as HSTL and SSTL to realize more than 100MHz operation. Also, Ramlink and SyncLink are briefly reviewed as candidates for next generation interface. Unified memory architecture attempts to save total memory capacity by combining graphics and main memory. Advanced device technology enables system integration which combine system logic and memory. It suggests one potential direction towards system on a chip in the future.

Ever increasing demand for higher bandwidth memories, which is fueled by multimedia and 3D graphics, seems to be somewhat satisfied with various emerging memory solutions. This paper gives a review of these emerging DRAM architectures and a performance comparison based on a condition to let the readers have some perspectives of the future and optimized graphics systems.

A single-chip architecture for three-dimensional (3-D) computer graphics (CG) is discussed assuming portable equipment with a 3-D CG interface. Based on a discussion of chip requirements, an architecture utilizing DRAM technology is proposed. A 31-Mbit, on-chip DRAM cell array allows a full-color, 480640-pixel frame with two 3-D frame buffers for double buffering and one 2-D frame buffer for superimposed or background images. The on-chip pixel generator produces R, G, B, and Z data in a triangular polygon with a zigzag-scan interpolation algorithm. The on-chip frame synthesizer combines data from one of the 3-D buffers with that from the 2-D buffer to produce superimposed or background 2-D images within a 3-D CG image. Parallel alpha-blending and Z-comparison circuits attached to the DRAM cell array provide a high data I/O rate. Estimation of the chip performance assuming the 0.35-µm CMOS design rule shows the chip size, the drawing speed, on-chip data I/O rate, and power dissipation would be 1413.5-mm, 0.25 million polygons/s, 1 gigabyte/s, and 590 mW at a voltage of 3.3 V, respectively. Based on circuit simulations, the chip can run on a 1.5-V dry cell with a drawing speed of 0.125 million polygons/s and a power dissipation of 61 mW. A scaled-down version of the chip which has an 1-kbit DRAM cell array with an attached alpha-blending circuit is being fabricated for evaluation.

In this paper we propose a new algorithm for recognizing 3-D objects from 2-D images. The algorithm takes the multiple view approach in which each 3-D object is modeled by a collection of 2-D projections from various viewing angles where each 2-D projection is called an object model. To select the candidates for the object model that has the best match with the input image, the proposed algorithm computes the surface matching score between the input image and each object model by using Hopfield nets. In addition, the algorithm gives the final matching error between the input image and each candidate model by the error of the pose-transform matrix proposed by Hong et al. and selects an object model with the smallest matching error as the best matched model. The proposed algorithm can be viewed as a combination of the algorithm of Lin et al. and the algorithm of Hong et al. However, the proposed algorithm is not a simple combination of these algorithms. While the algorithm of Lin et al. computes the surface matching score and the vertex matching score berween the input image and each object model to select the candidates for the best matched model, the proposed algorithm computes only the surface matching score. In addition, to enhance the accuracy of the surface matching score, the proposed algorithm uses two Hopfield nets. The first Hopfield net, which is the same as that used in the algorithm of Lin et al., performs a coarse matching between surfaces of an input image and surfaces of an object model. The second Hopfield net, which is the one newly proposed in this paper, establishes the surface correspondences using the compatibility measures between adjacent surface-pairs of the input image and the object model. the results of the experiments showed that the surface matching score obtained by the Hopfield net proposed in this paper is much more useful for the selectoin of the candidates for the best matched model than both the sruface matching score obtained by the first Hopfield net of Lin et al. and the vertex matching score obtained by the second Hopfield net of Lin et al. and, as the result, the object recognition algorithm of this paper can perform much more reliable object recognition than that obtained by simply combining the algorithm of Lin et al. and the algorithm of Hong et al.

This paper describes the architecture and new circuit technologies of a proposed Pixel (bit) -Aligned Triple-port DRAM (PATDRAM). The PATDRAM has a 270 K word 16 b Random Access Memory (RAM), a 512 word 8 b Serial Access Memory-(a) (SAMa) and a 1024 word 4 b Serial Access Memory-(b) (SAMb). The random port, serial-a and serial-b port can be operated by three independent synchronous clocks. In these three ports, word data can be aligned to the location of an arbitrary bit position. Data transfer from SAMb to RAM can be individually masked by transfer mask data. The RAM operates by 33 MHz synchronous clock and two SAMs operate by 40 MHz clocks. Novel architecture of the PATDRAM accelerates graphics performance and simplifies in multimedia systems which manage both realtime video and computer graphics data, and also accelerates graphics performance in both two-dimensional (2D) and three-dimensional (3D) graphics systems. PATDRAM was designed using a 0.6 µ double metal, triple poly, stacked capacitor, CMOS process technology in a 10.98 mm9.88 mm die area integrated 4.4 Mb RAM, 8 Kb SAM, 4 Kb transfer mask register and 5 Kgate logic.

The ray-tracing algorithm can synthesize very realistic images. However, the ray tracing is very time consuming. To solve this problem, a load balancing strategy using temporal coherence between images in an animation is presented for balancing computational loads among processing elements of a parallel processng system. Our parallel processing model is based on a space subdivision method for the ray-tracing algorithm. A subdivided object space is distributed among processing elements of the parallel system. To clarify the effectiveness of the load balancing strategy, we examine the system performance by computer simulation.

To derive blood flow dynamics from cineangiograms (CAG), we have developed an image processing algorithm to estimate a two-dimensional blood fiow velocity map projected on CAG. Each image area of CAG is diveded into blocks, and it is assumed that the movement of the contrast medium between two serial frames is restricted only to adjacent blocks. By this assumption, a fundamental equation" and the maximum flow constraints" are derived. The equation and constraints state the relationship between the volume of contrast medium in each block and the flow components" that are the volumes of contrast medium flowing from/to its adjacent blocks. The initial guess" that is a set of approximately obtained flow components is corrected using these relationships. The corrected flow components are then transformed into blood flow velocities, which are illustrated in the form of a needle diagram. In numerical experiments, the estimation error between the real flow velocity generated artificially and the flow velocity estimated with our algorithm was evaluated under one of the worst conditions. Although the maximum error was fairly large, the estimated flow velocity map was still acceptable for visual inspection of flow velocity pattern. We then applied our algorithm to an abdominal CAG (clinical data). The results showed flow stagnation and reverse flow in the abdominal aneurysm, which are consistent with the presence of a thrombus in the aneurysm. This algorithm may be a useful diagnostic tool in the assessment of vascular disease.

An algorithm interpolating parallel cross-sections between CT slices is described. Contours of equiscalar or constant-density surfaces on cross-sections are directly obtained as non-intersecting loops from grayscale slice images. This algorithm is based on a general algorithm that the authors have proposed earlier, constructing triangulated orientable closed surfaces from grayscale volumes and is particularly suited for a new technique, called laser stereolithography, which creates real 3D plastic objects using UV laser to scan and harden liquid polymer. The process of laser stereolithography is executed slice by slice, and this technique really requires some interpolation of intermediate cross-sections between slices. For visualizing, surfaces are only expected to be shaded almost continuously. The local defects are invisible and not cared about if the picture resolution is rather poor. On the contrary, topological faults are fatal to construct solid models by laser stereolithography, i.e., every contour line on cross-sections must be closed with no intersection. Not a single break of a contour line is tolerated. We already have many algorithms available for equiscalar surface construction, and it seems that if we cut the surfaces, then contour lines could be obtained. However, few of them are directly applicable to solid modeling. Marching cubes algorithm, for example, does not ensure the consistency of surface topology. Our algorithm guarantee an adequate topology of contour lines.

Surface construction is known as a way to visualize volume data. Although currently used algorithms such as marching cubes have good enough quality for volume visualization, they do not ensure adequate surface topology. These algorithms work well when the surface is rather simple. While when complicated, the surface does not separate the internal and external spaces, that is, there exist some holes on the surface, or exist redundant overlaps or self-intersection. Actually, adequate surface topology is important not only for visualization but for laser stereolithography, which creates real 3D plastic objects. In the present paper, we propose a new method that produces a set of triangular patches from a given volume data. The fact that the set of patches has no holes, no redundancy, no self-intersection, and has orientable closed surface topology is shown.

A process of mixing viscous fluids, such as oil-based paints is applied to generate marble patterns. It is difficult to get the exact flow function of the viscous fluid, then we express the flow in terms of velocity vectors derived from simplified flow phenomena, in which the viscous liquid is supposed to be a collection of finite liquid elements. The position change of each element is calculated as the function of time and several examples of the obtained marble patterns are illustrated.

Logarithmic number systems (LNS) provide a very fast computational method. Their exceptional speed has been demonstrated in signal processing and then in computer graphics. But the precision problem of LNS in computer graphics has not been fully examined. In this paper analysis is made for the problem of LNS in picture generation, in particular for circle drawing. Theoretical error analysis is made for the circle drawing. That is, some expressions are developed for the relative error variances. Then they are examined by simulation experiments. Some comparisons are also done with floating point arithmetic with equivalent word length and dynamic range. The results show that the theory and the experiments agree reasonably well and that the logarithmic arithmetic is superior to or at least comparable to the corresponding floating point arithmetic with equivalent word length and dynamic range. Those results are also verified by visual inspections of actually drawn circles. It also shows that the conversion error (from integer to LNS), which is inherent in computer graphics with LNS, does not make too much influence on the total computational error for circle drawing. But it shows that the square-rooting makes the larger influence.

A fast scan-line algorithm for a raster-scan graphics display is proposed based on an observation that a sequence of successive image frames in animation mostly consists of still objects with relatively few moving objects. In the proposed algorithm, successive images are generated using the background image composed of still objects only, and moving image composed only of moving objects. The color of each pixel in the successive images is then determined by one, which is nearer from eye, between the two candidate pixels, where one is from the background image and the other is from the moving image. The background image is generated once in the whole process, while the moving image is generated for each time frame using an interpolation of two images generated at the start and end time of the given time interval. For the purpose of fast shadow generation, we classify the shadows into three groups, i.e., still shadows generated by still objects on still objects, moving shadows generated by moving objects on still objects, and composite shadows generated by both still objects and moving objects on moving objects. These shadows can be generated very quickly by utilizing the frame coherence. According to the experimental results, a speed up factor of 3.2 to 12.8, depending on the percentage of the moving objects among all objects, was obtained using our algorithm, compared to the conventional scheme not utilizing the frame-to-frame image coherence.

New methods are needed for accessing very large information services. This paper proposes the use of a user model neural network to allow better access to a news service. The network is constructed on the basis of articles read, and articles marked as rejected. It adapts over time to better represent the user's interests and rank the articles supplied by the news service. Using an augmented keyword search we can also search for articles using keywords in conjunction with the user model neural network. Trials of the system in a USENET news environment show promising results for the use of this approach in information retrieval.

This paper presents an automatic creation method of 3D facial models which are needed for facial image generation by 3D computer graphics. A 3D facial model of a specific person is obtained from just the front and side view images without any human operation. The method has two parts; feature extraction and generic model modification. In the feature extraction part, the regions or edges which express the facial features such as eyes, nose, mouth or chin outline are extracted from the front and side view images. A generic head model is then modified based on the position and shape of the extracted facial features in the generic model modification part. As a result, a 3D model for persons is obtained. By using the specific model and the front and side view images, texture-mapped facial images can be generated easily.