Search

Radon transforms allow to represent n-dimensional objects by all their possible (n-1)-dimensional integrals. They find broad usage in a variety of image processing topics, covering pattern recognition and tomographic imaging. Potentially the most frequently used version is the 2D Radon transform which is commonly computed by means of the ray tracing procedure (Siddon, Joseph etc.).1 The problem comes down to a method of approximating a line integral through a 2D pixelized image. For higher dimensions, however, this problem becomes more and more complex and typically involves a substantial amount of case differentiation, making it particularly ill-suited for use in massively parallelized computation (e.g. on GPUs). Additionally, implementation effort is substantial and quite error-prone. Here, we propose a simple strategy to compute the (n-1)-dimensional integrals in a generalized manner by reducing the sampling problem to a single matrix multiplication. We further present OpenCL implementations for n=2 and n=3, making use of hardware interpolation methods on texture memory of GPU devices to provide a fast computation of the transform.

The SIFT algorithm is one of the most popular feature extraction methods and therefore widely used in all sort of video analysis tasks like instance search and duplicate/ near-duplicate detection. We present an efficient GPU implementation of the SIFT descriptor extraction algorithm using CUDA. The major steps of the algorithm are presented and for each step we describe how to efficiently parallelize it massively, how to take advantage of the unique capabilities of the GPU like shared memory / texture memory and how to avoid or minimize common GPU performance pitfalls. We compare the GPU implementation with the reference CPU implementation in terms of runtime and quality and achieve a speedup factor of approximately 3 - 5 for SD and 5 - 6 for Full HD video with respect to a multi-threaded CPU implementation, allowing us to run the SIFT descriptor extraction algorithm in real-time on SD video. Furthermore, quality tests show that the GPU implementation gives the same quality as the reference CPU implementation from the HessSIFT library. We further describe the benefits of GPU-accelerated SIFT descriptor calculation for video analysis applications such as near-duplicate video detection.

Reconstruction of 2 -d image primitives or of 3 -d volumetric primitives is one of the most common operationsperformed by the rendering components of modern visualization systems. Because this operation is often aided byGPUs, reconstruction is typically restricted to rst-order interpolation. With the advent of in situ visualization,the assumption that rendering algorithms are in general executed on GPUs is however no longer adequate. Wethus propose a framework that provides versatile texture ltering capabilities: up to third-order reconstructionusing various types of cubic ltering and interpolation primitives; cache-optimized algorithms that integrateseamlessly with GPGPU rendering or with software rendering that was optimized for cache-friendly \ Structureof Array " (SoA) access patterns; a memory management layer (MML) that gracefully hides the complexitiesof extra data copies necessary for memory access optimizations such as swizzling, for rendering on GPGPUs,or for reconstruction schemes that rely on pre- ltered data arrays. We prove the e ectiveness of our softwarearchitecture by integrating it into and validating it using the open source direct volume rendering (DVR) softwareDeskVOX.Keywords: Image Reconstruction, Splines, Texture Filtering Library, Direct Volume Rendering, HeterogeneousHPC Architectures

This paper compares the speed performance of a set of classic image algorithms for evaluating texture in images by using CUDA programming. We include a summary of the general program mode of CUDA. We select a set of texture algorithms, based on statistical analysis, that allow the use of repetitive functions, such as the Coocurrence Matrix, Haralick features and local binary patterns techniques. The memory allocation time between the host and device memory is not taken into account. The results of this approach show a comparison of the texture algorithms in terms of speed when executed on CPU and GPU processors. The comparison shows that the algorithms can be accelerated more than 40 times when implemented using CUDA environment.

As visualization is applied to larger data sets residing in more diverse hardware environments, visualization frameworks need to adapt. Rendering techniques are currently a major limiter since they tend to be built around central processing with all of the geometric data present. This is not a fundamental requirement of information visualization. This paper presents Abstract Rendering (AR), a technique for eliminating the centralization requirement while preserving some forms of interactivity. AR is based on the observation that pixels are fundamentally bins, and that rendering is essentially a binning process on a lattice of bins. By providing a more flexible binning process, the majority of rendering can be done with the geometric information stored out-of-core. Only the bin representations need to reside in memory. This approach enables: (1) rendering on large datasets without requiring large amounts of working memory, (2) novel and useful control over image composition, (3) a direct means of distributing the rendering task across processes, and (4) high-performance interaction techniques on large datasets. This paper introduces AR in a theoretical context, provides an overview of an implementation, and discusses how it has been applied to large-scale data visualization problems.

3D ultrasound computer tomography (3D USCT) promises reproducible high-resolution images for early detection of breast tumors. The synthetic aperture focusing technique (SAFT) used for image reconstruction is highly computeintensive but suitable for an accelerated execution on GPUs. In this paper we investigate how a previous implementation of the SAFT algorithm in CUDA C can be further accelerated and integrated into the existing MATLAB signal and image processing chain for 3D USCT. The focus is on an efficient preprocessing and preparation of data blocks in MATLAB as well as an improved utilisation of special hardware like the texture fetching units on GPUs. For 64 slices with 1024×1024 pixels each the overall runtime of the reconstruction including data loading and preprocessing could be decreased from 35 hours with CPU to 2.4 hours with eight GPUs.

This work extends the multi-histogram volume rendering framework proposed by Kniss et al. [1] to provide rendering results based on the impression of overlaid triangles on a graph of image intensity versus gradient magnitude. The developed method of volume rendering allows for greater emphasis to boundary visualization while avoiding issues common in medical image acquisition. For example, partial voluming effects in computed tomography and intensity inhomogeneity of similar tissue types in magnetic resonance imaging introduce pixel values that will not reflect differing tissue types when a standard transfer function is applied to an intensity histogram. This new framework uses developing technology to improve upon the Kniss multi-histogram framework by using Java, the GPU, and MIPAV, an open-source medical image processing application, to allow multi-histogram techniques to be widely disseminated. The OpenGL view aligned texture rendering approach suffered from performance setbacks, inaccessibility, and usability problems. Rendering results can now be interactively compared with other rendering frameworks, surfaces can now be extracted for use in other programs, and file formats that are widely used in the field of biomedical imaging can be visualized using this multi-histogram approach. OpenCL and GLSL are used to produce this new multi-histogram approach, leveraging texture memory on the graphics processing unit of desktops to provide a new interactive method for visualizing biomedical images. Performance results for this method are generated and qualitative rendering results are compared. The resulting framework provides the opportunity for further applications in medical imaging, both in volume rendering and in generic image processing.

This paper presents initial performance results for a software toolkit that implements GPU-based parallel computation of digitally reconstructed radiographs (DRRs) from volumetric imaging data for 2D-3D registration. The computational parallelism is achieved using NVIDIA’s CUDA implementation of general purpose computing on the graphics processing unit. The sample volumetric imaging data shown here is from CT imaging of a cadaveric foot, but the toolkit can be applied equally well to other volumetric imaging data. An efficient implementation requires launching hundreds of simultaneous, independent computational threads and fast thread access to the global memory where they need to read and write data. We have implemented fast DRR generation by launching a computational thread for each pixel in the image, and achieve efficient memory access by using 3D texture memory to store the volumetric data and constant memory to store global information such as intensifier coordinates. The Thrust software library was used to store individual bone DRRs, which enables efficient memory transfer and use of built-in device operators during image compositing and similarity quantification. By storing individual DRRs, the toolkit can support independent kinematics for up to 32 segmented objects. We show that the algorithm scales with the number of processors and compare timings for three commercially available GPUs. Here we present our initial fast DRR computations to demonstrate that the toolkit can produce useful results for a full 160 × 339 × 439 stack of floating point density data on a high resolution 1152 × 896 pixel screen in 1.3 ms and on a 512 × 512 pixel screen in less than 0.6 ms.

Performing high quality 3D visualizations on mobile devices, while tantalizingly close in many areas, is still a quite difficult task. This is especially true for 3D volume rendering of digital medical images. Allowing this would empower medical personnel a powerful tool to diagnose and treat patients and train the next generation of physicians. This research focuses on performing real time volume rendering of digital medical images on iOS devices using custom developed GPU shaders for orthogonal texture slicing. An interactive volume renderer was designed and developed with several new features including dynamic modification of render resolutions, an incremental render loop, a shader-based clipping algorithm to support OpenGL ES 2.0, and an internal backface culling algorithm for properly sorting rendered geometry with alpha blending. The application was developed using several application programming interfaces (APIs) such as OpenSceneGraph (OSG) as the primary graphics renderer coupled with iOS Cocoa Touch for user interaction, and DCMTK for DICOM I/O. The developed application rendered volume datasets over 450 slices up to 50-60 frames per second, depending on the specific model of the iOS device. All rendering is done locally on the device so no Internet connection is required.

In the photoelectrical tracking system, Bayer image is decompressed in traditional method, which is CPU-based. However, it is too slow when the images become large, for example, 2K×2K×16bit. In order to accelerate the Bayer image decoding, this paper introduces a parallel speedup method for NVIDA’s Graphics Processor Unit (GPU) which supports CUDA architecture. The decoding procedure can be divided into three parts: the first is serial part, the second is task-parallelism part, and the last is data-parallelism part including inverse quantization, inverse discrete wavelet transform (IDWT) as well as image post-processing part. For reducing the execution time, the task-parallelism part is optimized by OpenMP techniques. The data-parallelism part could advance its efficiency through executing on the GPU as CUDA parallel program. The optimization techniques include instruction optimization, shared memory access optimization, the access memory coalesced optimization and texture memory optimization. In particular, it can significantly speed up the IDWT by rewriting the 2D (Tow-dimensional) serial IDWT into 1D parallel IDWT. Through experimenting with 1K×1K×16bit Bayer image, data-parallelism part is 10 more times faster than CPU-based implementation. Finally, a CPU+GPU heterogeneous decompression system was designed. The experimental result shows that it could achieve 3 to 5 times speed increase compared to the CPU serial method.

Compared to iris recognition, sclera recognition which uses line descriptor can achieve comparable recognition accuracy in visible wavelengths. However, this method is too time-consuming to be implemented in a real-time system. In this paper, we propose a GPU-based parallel computing approach to reduce the sclera recognition time. We define a new descriptor in which the information of KD tree structure and sclera edge are added. Registration and matching task is divided into subtasks in various sizes according to their computation complexities. Every affine transform parameters are generated by searching on KD tree. Texture memory, constant memory, and shared memory are used to store templates and transform matrixes. The experiment results show that the proposed method executed on GPU can dramatically improve the sclera matching speed in hundreds of times without accuracy decreasing.

In computer graphics, rendering algorithms are used to simulate the appearance of objects and materials in a wide range of applications. Designers and manufacturers rely entirely on these rendered images to previsualize scenes and products before manufacturing them. They need to differentiate between different types of fabrics, paint finishes, plastics, and metals, often with subtle differences, for example, between silk and nylon, formaica and wood. Thus, these applications need predictive algorithms that can produce high-fidelity images that enable such subtle material discrimination.

Iterative CT reconstruction methods have advantages over analytical reconstruction methods because of their robustness to both noise and incomplete projection data, which have great potential for dose reduction in real applications. The SART algorithm, which is one of the well-established iterative reconstruction methods, has been examined extensively, and GPU has been applied to improve their efficiency. Although it has been proved that SART may globally converge, its convergence is very slow, especially after the first several iterations. Hundreds of iterations may be needed for accurate reconstruction. This slow convergence requires heavy data transfer between global memory and texture memory inside GPU. Therefore, preconditioned conjugate gradient (CG) method, which converges much faster than SART, may be combined with SART for better performance. Since CG is sensitive to initialization, the reconstruction results from SART after a few iterations may be used as the initialization for CG. Preliminary experimental results on CPU show that this framework converges much faster than SART and CG, which demonstrates its potential in real applications.

This paper presents an efficient graphics hardware-based method to segment and visualize level-set surfaces as interactive rates. Our method is composed of page manager, level-set solver, and volume renderer. The page manager which performs in CPU generates page table, inverse page table and available page stack as well as processes the activation and inactivation of pages. The level-set solver computes only voxels near the iso-surface. To run efficiently on GPUs, volume is decomposed into a set of small pages. Only those pages with non-zero derivatives are stored on GPU. These active pages are packed into a large 2D texture memory. The level-set partial differential equation (PDE) is computed directly on this packed format. The page manager is used to help managing the packing of the active data. The volume renderer performs volume rendering of the original data simultaneously with the evolving level set in GPU. Experimental results using two chest CT datasets show that our graphics hardware-based level-set method is much faster than software-based one.

The inspection of a patient's data for diagnostics, therapy planning or therapy guidance involves an increasing number of 3D data sets, e.g. acquired by different imaging modalities, with different scanner settings or at different times. To enable viewing of the data in one consistent anatomical context fused interactive renderings of multiple 3D data sets are desirable. However, interactive fused rendering of typical medical data sets using standard computing hardware remains a challenge. In this paper we present a method to render multiple 3D data sets. By introducing local rendering functions, i.e. functions that are adapted to the complexity of the visible data contained in the different regions of a scene, we can ensure that the overall performance for fused rendering of multiple data sets depends on the actual amount of visible data. This is in contrast to other approaches where the performance depends mainly on the number of rendered data sets. We integrate the method into a streaming rendering architecture with brick-based data representations of the volume data. This enables efficient handling of data sets that do not fit into the graphics board memory and a good utilization of the texture caches. Furthermore, transfer and rendering of volume data that does not contribute to the final image can be avoided. We illustrate the benefits of our method by experiments with clinical data.

Volume rendering has high utility in visualization of segmented datasets. However, volume rendering of the segmented labels along with the original data causes undesirable intermixing/bleeding artifacts arising from interpolation at the sharp boundaries. This issue is further amplified in 3D textures based volume rendering due to the inaccessibility of the interpolation stage. We present an approach which helps minimize intermixing artifacts while maintaining the high performance of 3D texture based volume rendering - both of which are critical for intra-operative visualization. Our approach uses a 2D transfer function based classification scheme where label distinction is achieved through an encoding that generates unique gradient values for labels. This helps ensure that labelled voxels always map to distinct regions in the 2D transfer function, irrespective of interpolation. In contrast to previously reported algorithms, our algorithm does not require multiple passes for rendering and supports greater than 4 masks. It also allows for real-time modification of the colors/opacities of the segmented structures along with the original data. Additionally, these capabilities are available with minimal texture memory requirements amongst comparable algorithms. Results are presented on clinical and phantom data.

Direct volumetric visualization of medical datasets has important application in areas such as minimally-invasive therapies and surgical simulation. In popular fixed-slice-distance hardware-based volume rendering algorithms, such as 2D and 3D texture mapping, the non-isotropic nature of the volumetric medical images and the constantly changed viewing rays make it difficult to render medical datasets without disturbing or slicing artifacts during volume rotation. We have developed a hardware accelerated 3D medical image visualization system based on a commodity graphics unit, in which a viewing-direction based dynamic texture slice resampling scheme is descirbed and implemented on an Nvidia graphics processing unit (GPU). In our algorithm, we utilize graphics hardware to dynamically slice the volume texture according to the viewing directions during the rendering process, in which the slice number can be dynamically changed without consuming additional video memory. Near-uniform effective slice spacing can be achieved in real-time and updated as the viewing angles change, so improved uniform visual quality is achieved with high rendering performance. To further improve rendering efficiency, we have implemented a multi-resolution scheme within our rendering system, which offers the user the option to highlight the volume of interest (VOI) and render it with higher resolution than the surrounding structures. This system also incorporates a fragment-level interactive post-classification algorithm that modifies the texture directly within the texture unit on graphics card, making it possible to interactively change transfer function parameters and navigate medical datasets in real-time during the 3D medical image visualization process.

Modern optical imaging techniques such as confocal and multi-photon microscopy can acquire volumetric datasets of cellular structures. In this paper we propose an approach for interactive volume rendering of such cellular datasets. In the first stage, we create a set of 2D textures corresponding to the image stacks in the original dataset. These textures are generated through a transfer function that maps voxel intensities to colors and opacities, and stored in the texture memory in computer. In the second stage, by blending the textures with hardware support, we can achieve interactive volume rendering including rotation and zooming on regular PCs. Besides, to generate good images for viewing in lateral directions, we use two additional sets of 2D textures for two orthogonal lateral directions and the texture resolutions can be adapted to the rendering requirement and computer hardware. This approach offers an effective visualization environment for biologists to better understand and analyze measured cellular structures.

The pre-integrated volume rendering which produces high-quality images with less sampling has become one of the most efficient and important techniques in volume rendering field. In this paper, we propose an acceleration technique of pre-integrated rendering of dynamically classified volumes. Using the overlapped-min-max block, empty space skipping of ray casting can be applied in pre-integrated volume rendering. In addition, a new pre-integrated lookup table brings much fast rendering of high-precision data without degrading image quality. We have implemented our approaches not only on the consumer graphics hardware but also on CPU, and show the performance gains using several medical data sets.

In this paper, we propose a novel technique of multimodality volume fusion using a graphics hardware. Our 3D texture based volume fusion algorithm consists of three steps: First, two volumes of different modalities are loaded into the texture memory in the GPU. Second, textured slices of two volumes along the same proxy geometry are combined with various compositing functions. Third, all the composited slices are alpha blended. We have implemented our algorithm using HLSL (High Level Shader Language). Our method shows that the exact depth of each volume and the realistic views with interactive rate in comparison with the software-based image integration. Experimental results using MR and PET brain images and the angiography with a stent show that over composting operation is more useful for clinical application.

Previously shell rendering was shown to be an ultra fast method of rendering digital surfaces represented as sets of voxels. The purpose of this work is to describe a new extension to the shell rendering method that creates a new representation of surfaces called t-shells using triangulated surface elements. The great speed of the shell rendering technique is made available to rendering triangulated surfaces through the use of data structures that describe the shell and through algorithms that traverse this information to produce the two-dimensional projection of the three-dimensional data. In traditional shell rendering, each shell element (a voxel) is represented by a triple comprised of the offset of the voxel from the start of the row, the neighborhood code of the voxel, and the surface normal. We modify this data structure by replacing the neighborhood code with a code that indicates the Marching Cubes configuration of polygons (1 of 256 possible triangulated configurations) within that area (rather than the rasterization of a single, uniform shell element as was originally done). We present the general t-shell algorithm as well as the results of the two implementations as applied to input data which consist of different objects from various parts of the body and various modalities with a variety of surface sizes and shapes. We present some sample renditions as well as the results of timing experiments which demonstrate that the t-shell rendering is not only much faster than hardware-based rendering but is also capable of handling scenes that are much larger than those the hardware is capable of rendering while producing renditions of comparable quality.

Northrop Grumman Amherst Systems has continued to improve the Real-time IR/EO Scene Simulator (RISS) for hardware-in-the-loop (HWIL) testing of infrared sensor systems. Several new and enhanced capabilities have been added to the system for both customer and internal development programs. A new external control capability provides control of either player trajectories or unit-under-test (UUT) orientation. The RISS Scene Rendering Subsystem (SRS) has been enhanced with support for texture transparency and increased texture memory capacity. The RISS Universal Programmable Interface (UPI) graphical user interface (GUI) has been improved to provide added flexibility and control of the real-time sensor modeling capabilities. This paper will further explore these and other product improvements.

A revolution is underway within commercial PC video graphics, driven mainly by the 3-D gaming community and its demands for customizable lighting effects and realistic, visually appealing, 3-D rendering. This revolution is bringing about a configurable transformation and lighting (T&L) engine within modern PC video graphics hardware. The results of these technological advancements will profoundly impact the way computer-based rendering is done. Although PC graphics hardware continues to change rapidly, it has evolved to the point where it can be made to address most of the Hardware-In-the-Loop (HWIL) scene generation demands which historically could be accomplished only on costly graphics workstations. With the ability to control how operations are performed within the hardware rendering process, it is possible to implement customized per-pixel spatial and lighting effects. To illustrate how these capabilities can be applied to solve certain HWIL scene generation problems. A graphics hardware approach will be implemented to demonstrate a method of achieving increased monochrome intensity resolution and a user-defined spatial distortion. There is great potential in modern graphics hardware. The limits are becoming less a function of the hardware capabilities and more a function of the ability of engineers and scientists to exploit the functionality of this rapidly advancing hardware rendering technology.

In Medical Imaging, shell rendering and shear-warp rendering are two of the most efficient and effective voxel-based techniques for volume visualization. This work presents a comparative analysis of shell rendering and shear-warp rendering in terms of storage, speed, and image quality. We have chosen 10 different objects of various sizes, shapes and topologies and one 1-GHz Pentium-III PC with 512 MB RAM for our experiments. Hard and fuzzy boundaries of up to 2,833 K voxels in size have been created to test both methods in surface and volume rendering, respectively. Hard surface shell rendering and surface shear-warp rendering required less than 0.5 second. In the worst case, volume shell rendering required 1.45 second, while volume shear-warp rendering spent 0.65 second for the same task. Shear-warp rendering uses on average from 3 to 6 times more memory space than shell rendering, but it can be up to 2.79 times faster than shell rendering. On average, shear-warp rendering is as fast as shell rendering for hard boundaries and 1.7 times faster than shell rendering for fuzzy boundaries. We have also observed that both can produce similar high-quality images.

In computerized image and graphic applications, texture mapping is one of the most commonly used methods to improve the realism or to enhance the visual effect of object rendering without too much increase in computational complexity. In conventional texture mapping, a three-dimensional (3D) object has to be transferred to the polygonal structure, and then mapped with a texture or photographic image. However, it is usually computationally expensive to transfer the original data in voxels to polygons, and is even more complex to map the 2D texture image onto the 3D polygonal structure. Even though the computation is huge, the polygonal structure is not efficient in preserving the internal information of volume data. Because most volume data acquired by medical imaging devices or 3D scanners are in voxel format, it is more appropriate to handle these data directly in voxel format. In this paper, we propose a new texture mapping method based on chain-coding flattening to handle the voxel-based data directly. Therefore, the computation is reduced significantly and the internal information can be utilized and preserved thoroughly. The method flattens a 3D object surface onto a 2D plane and then uses 2D warping technique to generate the correspondences between the object surface and the texture image. Therefore, polygonal transformation required in the conventional approach is no longer necessary and texture mapping is handled with inexpensive 2D computation. Experimental results have shown the effectiveness and efficiency of the proposed algorithm.

Recent advances in real-time synthetic scene generation for Hardware-in-the-loop (HWIL) testing at the U.S. Army Aviation and Missile Command (AMCOM) Aviation and Missile Research, Development, and Engineering Center (AMRDEC) improve both performance and fidelity. Modeling ground target scenarios requires tradeoffs because of limited texture memory for imagery and limited main memory for elevation data. High- resolution insets have been used in the past to provide better fidelity in specific areas, such as in the neighborhood of a target. Improvements for ground scenarios include smooth transitions for high-resolution insets to reduce high spatial frequency artifacts at the borders of the inset regions and dynamic terrain paging to support large area databases. Transport lag through the scene generation system, including sensor emulation and interface components, has been dealt with in the past through the use of sub-window extraction from oversize scenes. This compensates for spatial effects of transport lag but not temporal effects. A new system has been developed and used successfully to compensate for a flashing coded beacon in the scene. Other techniques have been developed to synchronize the scene generator with the seeker under test (SUT) and to model atmospheric effects, sensor optic and electronics, and angular emissivity attenuation.

In this paper, we survey the techniques for image-based rendering. Unlike traditional 3D computer graphics in which 3D geometry of the scene is known, image-based rendering techniques render novel views directly from input images. Previous image-based rendering techniques can be classified into three categories according to how much geometric information is used: rendering without geometry, rendering with implicit geometry (i.e., correspondence), and rendering with explicit geometry (either with approximate or accurate geometry). We discuss the characteristics of these categories and their representative methods. The continuum between images and geometry used in image-based rendering techniques suggests that image-based rendering with traditional 3D graphics can be united in a joint image and geometry space.

At the 1999 meeting of SPIE Medical Imaging, we presented results that demonstrated that a software-only-based technique, Shell Rendering, was capable of rendering hard surfaces at rates faster than hardware-assisted, triangle- based surface rendering by a factor of 18 to 31 while maintaining comparable image quality. We noted that the framework of Shell Rendering actually encompasses both hard surface rendering of medical image data via Shell Rendering in software with two techniques of OpenGL-based hardware- assisted volume rendering using 2D and 3D texture mapping. Although our previous results demonstrated that shell rendering is faster than hardware-assisted surface rendering, its speed is comparable to but slightly lower than hardware-assisted volume rendering methods. Detailed timing results for various input medical image data sets as well as for various computer platforms are presented in the paper. The images produced by the various methods are presented as well for subjective quality assessment. We conclude that for medical image visualization, Shell Rendering is an efficient technique that combines aspects of surface rendering and volume rendering. Furthermore, it accomplishes this without requiring expensive hardware while producing high quality images on PCs using an entirely software-based approach.

The purpose of this work is to compare the speed of isosurface rendering in software with that using dedicated hardware. Input data consists of 10 different objects form various parts of the body and various modalities with a variety of surface sizes and shapes. The software rendering technique consists of a particular method of voxel-based surface rendering, called shell rendering. The hardware method is OpenGL-based and uses the surfaces constructed from our implementation of the 'Marching Cubes' algorithm. The hardware environment consists of a variety of platforms including a Sun Ultra I with a Creator3D graphics card and a Silicon Graphics Reality Engine II, both with polygon rendering hardware, and a 300Mhz Pentium PC. The results indicate that the software method was 18 to 31 times faster than any hardware rendering methods. This work demonstrates that a software implementation of a particular rendering algorithm can outperform dedicated hardware. We conclude that for medical surface visualization, expensive dedicated hardware engines are not required. More importantly, available software algorithms on a 300Mhz Pentium PC outperform the speed of rendering via hardware engines by a factor of 18 to 31.

A graphics-based simulation has been proved as a powerfiul tool in surgical and therapeutic applications. In a computer aidedsurgical simulation system, it must provide real-time rendering performance as well as flexible 3D object manipulation. In thispaper, we propose a new volume rendering scheme, namely SSB (Scan-line Based Semi-Boundary). The time complexity ofSSB is 0 ( MN) plus a 2-Dimensional image rotation, and the SSB is theoretically faster than the original Semi-Boundaryalgorithm that will take 0 ( LMN ) time, where L , M , N are the Semi-Boundary nodes in X, Y, Z axes, respectively. Weexperimentally evaluate SSB rendering performance using large pelvis volume data. Our preliminary results show that SSB can achieve an interactive rendering performance for large volume pelvis data set with size of 5 1 2x5 12x245. Key Words : SSB, SB, semi boundary, Surgical Simulation , Volume Rendering , Pelvis. 1. INTRODUCTION 3 D visualization technique has been exploited to explore new discovery of science in many areas such as engineering simulation

Compression in the context of texture and bump mapping brings high-performance texture mapping in the range of low- cost systems. The size for the texture memory is significantly reduced and the necessary bandwidth between the memory and texturing unit is lowered. Textures are compressed using a lossy compression method based on vector quantization. We are proposing an alternative approach to embossing textured surfaces, where bump maps are described with bitmaps and are part of the compressed texture data. The design of a texturing circuit architecture is presented, where embossing or engraving of textured surfaces is executed by dedicated decoding and filtering hardware. The circuit decompresses textures and decodes compressed bump maps to produce wrinkled textured surfaces. Applying encoding to normal vectors results in narrower data paths. Vectors are represented and handled in a compressed format defined by their vertical and horizontal angles. In order to enhance subjective image quality, an optional space-variant filter can be applied locally during the texture mapping phase to reduce the artifacts introduced by the lossy compression method.© (1997) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

A general framework for fast visualization of multispectral volume data is presented. Dedicated hardware with a non-numeric coprocessor is utilized in the first step of the rendering pipeline to process the volume data and extract voxels according to feature characteristics. This capability is used to select voxels according to automatic classification results or real-time descriptions of regions of interest supplied by the user in an interactive environment. By this step we can in real-time reduce the number of voxels that have to be considered in the rendering and increase the speed of the volume rendering accordingly. The selected voxels are generated in a front-to-back (or back-to-front) order and projected to the view plane where a 3D rendering is accumulated with an adaptation of the shell rendering technique proposed by Udupa and Odhner. The paper includes an overview of the underlying hardware architecture and presents numerical experiments with a software simulator.© (1995) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

A method is given for synthesizing a texture by using the interface of a conventional drawing tool. The majority of conventional texture generation methods are based on the procedural approach, and can generate a variety of textures that are adequate for generating a realistic image. But it is hard for a user to imagine what kind of texture will be generated simply by looking at its parameters. Furthermore, it is difficult to design a new texture freely without a knowledge of all the procedures for texture generation. Our method offers a solution to these problems, and has the following four merits: First, a variety of textures can be obtained by combining a set of feature lines and attribute functions. Second, data definitions are flexible. Third, the user can preview a texture together with its feature lines. Fourth, people can design their own textures interactively and freely by using the interface of a conventional drawing tool. For users who want to build this texture generation method into their own programs, we also give the language specifications for generating a texture. This method can interactively provide a variety of textures, and can also be used for typographic design.© (1995) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

The image reconstruction problem, also known as the inverse Radon transform, for x-ray computed tomography (CT) is found in numerous applications in medicine and industry. The most common algorithm used in these cases is filtered backprojection (FBP), which, while a simple procedure, is time-consuming for large images on any type of computational engine. Specially designed, dedicated parallel processors are commonly used in medical CT scanners, whose results are then passed to a graphics workstation for rendering and analysis. However, a fast direct FBP algorithm can be implemented on modern texture-mapping hardware in current high-end workstation platforms. This is done by casting the FBP algorithm as an image warping operation with summing. Texture- mapping hardware, such as that on the silicon Graphics Reality Engine, shows around 600 times speedup of backprojection over a CPU-based implementation (a 100 Mhz R4400 in our case). This technique has the further advantages of flexibility and rapid programming. In addition, the same hardware can be used for both image reconstruction and for volumetric rendering. Our technique can also be used to accelerate iterative reconstruction algorithms. The hardware architecture also allows more complex operations than straight-ray backprojection if they are required, including fan-beam, cone-beam, and curved ray paths, with little or no speed penalties.© (1994) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Most texture mapping involves a backward projection which maps from screen space back into the texture data. Bilinear/trilinear interpolation is usually used to sample data in the 2D/3D textures. To accelerate texture mapping, traditional graphics hardware requires a complete copy of the texture image at each parallel computation node to allow all nodes to operate in parallel. This paper introduces a new backward projection approach for rendering volume data, treating the volume as a 3D texture. It employs an innovative parallelization scheme in which a computation node operates on a subvolume, thereby avoiding the need to store the whole volume at each node. This approach has been implemented on the Kubota Kenai Denali workstation, a high performance system for imaging and 3D graphics. It currently supports functions for maximum intensity projection, multiplanar reformatting, volume resampling, ray sum, and isosurface rendering. Progressive refinement schemes can be utilized to reduce sampling rate and to realize interactive rendering speeds.© (1994) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

We present a new paradigm, called shell rendering, for volume visualization of surfaces. It significantly overcomes the two major impediments of current volume rendering methods -- high computational cost and storage requirements -- and makes interactive volume rendering feasible in a workstation environment via portable software. It imparts the notion of a structure -- a shell -- to what is usually treated as an amorphous semi-transparent volume and makes morphometrics of surfaces feasible in the volume rendering paradigm.© (1992) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

The advent of `photorealistic' rendering has caused a great increase in the use of discrete texture mapping for image synthesis. Strategies to avoid the introduction of visible artifacts in texture mapping are varied. Most often, texture filtering is at the heart of these strategies. A particularly difficult problem to solve efficiently arises when a filtering operation varies significantly from pixel to pixel because it is often not clear how to perform preprocessing, the cost of which would be amortized over many pixels. We call such filtering operations space- variant filtering. In this paper, we describe a variety of such techniques, and we describe our recent attempts to develop constant-time implementations of space-variant filtering operations.© (1992) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.