Tree modeling and rendering is an integral part of many modern computergraphics applications. Unfortunately, tree models are highly detailed and require a lot of geometry information. To solve this problem bothimage...
详细信息
ISBN:
(纸本)0889865981
Tree modeling and rendering is an integral part of many modern computergraphics applications. Unfortunately, tree models are highly detailed and require a lot of geometry information. To solve this problem bothimage- and geometry-based simplification techniques have been proposed. these techniques build multiresolution representations that either have large storage requirements or do not allow viewing the trees from a close distance. We present a multiresolution model that can be efficiently stored in the GPU and produces highly realistic views of trees at close range. Our model supports a rendering algorithm that only requires two render operations to display any level of detail of any tree. We propose a simplification method targeted at this rendering algorithm. this simplification method produces continuous levels of detail of the leaves of each tree. We show that our algorithm can render scenes with as many as several million trees modeled using our representation.
In this paper, interest is focused on the accurate and time-efficient stereo reconstruction, for the purpose of generating 3D animated scenes from multiple synchronized videos. the plane-sweeping approach is reviewed ...
详细信息
ISBN:
(纸本)9781424404810
In this paper, interest is focused on the accurate and time-efficient stereo reconstruction, for the purpose of generating 3D animated scenes from multiple synchronized videos. the plane-sweeping approach is reviewed as relevant to the goal of time-efficiency, since its execution can be optimized on a GPU. A method compatible for optimization on the GPU is proposed as a more accurate alternative to plane sweeping and to the derived visibility computation. the method is compared to plane sweeping as to its accuracy, by evaluating the backprojected 3D model against independent views and using n-fold cross validation to estimate the Peak Signal to Noise Ratio (PSNR). Finally, the method's output is casted integratable with multicamera stereo reconstruction frameworks.
In this paper we propose a combination of different Haar filter sets and Edge Orientation Histograms (EOH) in order to learn a model for pedestrian detection. As we will show, withthe addition of EOH we obtain better...
详细信息
ISBN:
(纸本)0889865981
In this paper we propose a combination of different Haar filter sets and Edge Orientation Histograms (EOH) in order to learn a model for pedestrian detection. As we will show, withthe addition of EOH we obtain better ROCs than using Haar filters alone. Hence, a model consisting of discriminant features, selected by AdaBoost, is applied at pedestrian-sized image windows in order to perform the classification. Additionally, taking into account the final application, a driver assistance system with realtime requirements, we propose a novel stereo-based camera pitch estimation to reduce the number of explored windows. Withthis approach, the system can work in urban roads, as will be illustrated by current results.
this work presents a method to construct featured 3D terrains using information gathered from their satellites images. this data is used as keys in an image-database search. these searches use wavelets for minimize di...
详细信息
ISBN:
(纸本)0889865981
this work presents a method to construct featured 3D terrains using information gathered from their satellites images. this data is used as keys in an image-database search. these searches use wavelets for minimize disk space and improve searching speed. In order to compose the elements with trees or bushes, for example, we use the common billboard concepts embedded with transparent technique named alpha-channel.
Log-Polar transform is mentioned as one of the models of the central-fovea visual sensor in a creature sight. Up to now, in the field of image recognition, Log-Polar transform is known widely and is especially used we...
详细信息
ISBN:
(纸本)0889865981
Log-Polar transform is mentioned as one of the models of the central-fovea visual sensor in a creature sight. Up to now, in the field of image recognition, Log-Polar transform is known widely and is especially used well in recognition of face images. However, since it is the recognition method eternal only to a scaling and a rotation, it mainly remains in application in a 2D space. In this paper, a new Log-Polar transform algorithm which can detect even affine transform parameters is proposed. By using this transform, the pattern recognition within 3D space is possible.
Algorithms for imageprocessing and computervision are natural candidates for high performance computing systems. this paper presents a reconfigurable parallel architecture prototype for imageprocessing base on larg...
详细信息
ISBN:
(纸本)1424403316
Algorithms for imageprocessing and computervision are natural candidates for high performance computing systems. this paper presents a reconfigurable parallel architecture prototype for imageprocessing base on large scale FPGA computing. the introduced architecture can cover a wide range of real-time computervision applications from pre-processing operations to low-level interpretation. In order to reduce the memory accessing time and communication latency, prime memory system for neighborhood operations or other data structures and FPGA-based transfer interconnection networks were designed in the introduced prototype system. this proposed architecture allows the user to program the system in both high-lever (soft-programming) and low-level (hard-programming).
Traditional computervision applications are often challenged by the need to distinguish between objects and their shadows. though, many shadow detection and removal algorithms have been proposed in literature. Most o...
详细信息
ISBN:
(纸本)0889865981
Traditional computervision applications are often challenged by the need to distinguish between objects and their shadows. though, many shadow detection and removal algorithms have been proposed in literature. Most of the proposed methods that claim to be object and environment independent include some minor assumptions about the scene geometry or spectral distributions of the light sources. In this work, based on a physically-derived hypothesis for shadow identification, novel, simple and fast shadow detection algorithms are proposed and implemented in the spatial (Pixel) and frequency (Fourier) domains. It is shown that the algorithms effectively remove shadows under various lighting and environmental conditions. the proposed algorithms are able to detect shadows in both umbra and penumbra neighbourhoods.
Non-supervised multiple-agent tracking is a complex task which demands a structured framework in order to accomplish it. therefore, this proposal presents a system which is modular and hierarchically organised. It con...
详细信息
ISBN:
(纸本)0889865981
Non-supervised multiple-agent tracking is a complex task which demands a structured framework in order to accomplish it. therefore, this proposal presents a system which is modular and hierarchically organised. It consists in several levels, working in cascade, which are defined according to the different functionalities to be performed. the goal of this work is to implement and experimentally verify a novel image-based algorithm which deals with serious segmentation difficulties, thereby being able to simultaneously perform a reliable tracking of several agents. As a result, agents' trajectories are obtained, as well as quantitative information about their state at any time, such as their speed or size.
this paper introduces an idea of using a PMD sensor to implement a feature-based stereo algorithms in a cost efficient way. Photonic Mixer Device (PMD) is a semiconductor sensor that has been developed for 3D reconstr...
详细信息
ISBN:
(纸本)0889865981
this paper introduces an idea of using a PMD sensor to implement a feature-based stereo algorithms in a cost efficient way. Photonic Mixer Device (PMD) is a semiconductor sensor that has been developed for 3D reconstruction of the scene. A common approach to implement a feature-based stereo algorithm in a cost efficient way is to introduce constraints which reduce the search area. this paper introduces a constraint "initial disparity guess" calculated by using the depth information provided by a PMD. this allows searching a specific feature point of a stereo image in the shorter range of the scan line of corresponding image. the initial results and findings of the proposed system are shown for the 3D reconstruction of a simple object using real data acquired from both PMD sensor and stereo cameras.
In this paper we present a method to generate a digital mosaic starting from a raster input image. Mosaics generation of artistic quality is challenging. the basic elements, the tiles, typically small polygons, must b...
详细信息
暂无评论