This work presents an implementation of a hierarchical disparity estimation algorithm entirely executed on programmable 3D graphics hardware. In contrast to previous hardware based implementations of computational stereo algorithms, our method calculates disparities either for rectified stereo images or uncalibrated pairs of stereo images without known epipolar geometry. We exploit features of modern graphics hardware to search for correct disparity vectors e#ciently. The hierarchical approach increases the speed and the robustness of the algorithm. Additionally, we use bidirectional matching to remove false matches. We observe up to 50 fps for input images with 256 256 pixels resolution on desktop graphics cards and up to 30 fps on mobile 3D hardware.

We present an accurate, interactive silhouette extraction mechanism for texture-based volume rendering. Compared to previous approaches, our system guarantees silhouettes of a user controlled width without any significant preprocessing time. Our visualization pipeline consists of two steps: (a) extraction of silhouettes with a width of one pixel, and (b) image post-processing for broadening of silhouettes. Step (a) is a mixture of object- and image-based- silhouette extraction models, maximally exploiting the screen resolution. This hybrid approach is neither sensitive to accuracy in gradient representation nor to the precision of the depth-buffer, as in earlier procedures. Step (b) is accomplished via smoothing and applying a threshold to the temporary result obtained in (a). To keep the latter process efficient, we perform fast convolution using FFT. Our silhouette extraction is conceptually similar to the corresponding method for polygonal representations, checking the front-and back facing property of adjacent triangles.

This paper presents a 3D, volumetric, seismic fault detection system that relies on a novel set of nonlinear filters combined with a GPU (Graphics Processing Unit) implementation, which makes interactive nonlinear, volumetric processing feasible. The method uses a 3D structure tensor to robustly measure seismic orientations. These tensors guide an anisotropic diffusion, which reduces noise in the data while enhancing the fault discontinuity and coherency along seismic strata. A fault-likelihood volume is computed using a directional variance measure, and 3D fault voxels are then extracted through a non-maximal-suppression method. We also show how the proposed algorithms are efficiently implemented with a GPU programming model, and compare this to a CPU implementation to show the benefits of using the GPU for this computationally demanding problem.

Abstract. Denoising video is an important task, especially for videos captured in dim lighting environments. The filtering of video in a volumetric manner with time as the third dimension can improve the results significantly. In this work a 3D bilateral filter for edge preserving smoothing of video sequences exploiting commodity graphics hardware is presented. A hardware friendly streaming concept has been implemented to allow the processing of video sequences of arbitrary length. The clear advantage of time-based filtering compared to frame-by-frame filtering is presented as well as solutions to current limitations for volume filtering on graphics hardware. In addition, a significant speedup over a CPU based implementation is shown. Keywords: Non-Linear Filtering, Graphics Hardware, Video Processing. 1

Abstract not found

The ultimate goal of many applications of augmented reality is to immerse the user into the augmented scene, which is enriched with virtual models. In order to achieve this immersion, it is necessary to create the visual impression that the graphical objects are a natural part of the user's environment. Producing this effect with conventional computer graphics algorithms is a complex task. Various rendering artifacts in the three-dimensional graphics create a noticeable visual discrepancy between the real background image and virtual objects.

Figure 1: In the left three images, an interactive segmentation of a brain tumor with an active surface model given in implicit form is evolving toward the final segmentation using the level set method [Lefohn et al. 2003]. In the right image, the same method is applied to segmentation of a mouse liver. Level set computations and simultaneous visualization are performed entirely on the GPU (graphics processing unit), which performs general floating-point computations in addition to rendering. Recent advances in the computational power of graphics processing units (GPUs) have turned them into a viable platform for general purpose floating-point computations. A very promising application of these new capabilities is interactive segmentation of medical volume data, which usually involves solving a large number of partial differential equations (PDEs) for each iteration of an evolving segmentation that can be viewed and guided by user input while it is being calculated, and is thus computationally very demanding. We give an overview of segmentation algorithms with a focus on leveraging the power of GPUs in order to obtain high-quality segmentations of medical data in an interactive process, with the premise that these algorithms will lead to faster and higher-quality segmentations in clinical practice in the near future. 2

Programmable graphics processing units (GPUs) are hardware commonly included in new computer workstations, including those workstations used for digital signal and image processing. The current generation of GPU is roughly equivalent in

Minimally invasive image-guided interventions (IGIs) are time and cost efficient, minimize unintended damage to healthy tissues, and lead to faster patient recovery. Advanced three-dimensional (3D) image processing is a critical need for navigation during IGIs. However, achieving on-demand performance, as required by IGIs, for these image processing operations using software-only implementations is challenging because of the sheer size of the 3D images, and memory and compute intensive nature of the operations. This dissertation, therefore, is geared toward developing high-performance 3D image processing architectures, which will enable improved intraprocedural visualization and navigation capabilities during IGIs. In this dissertation we present an architecture for real-time implementation of 3D filtering operations that are commonly employed for preprocessing of medical images. This architecture is approximately two orders of magnitude faster thancorresponding software implementations and is capable of processing 3D medical images at their acquisition speeds. Combining complementary information through registration between pre- and intraprocedural images is a fundamental need in the IGI workflow. Intensity-based

Abstract not found