【摘要】 在大规模图像可视化应用中,经过四叉树预处理,建立多分辨率图像金字塔索引,以支持对不同分辨率图像的快速读取及显示。但是在显示图像分辨率与金字塔中保存图像分辨率不一致情况下,依然需要通过图像重采样来获得指定分辨率图像。一般重采样过程可以通过基于CPU或者GPU的重采样方法来完成。论文通过建立RECG(resampling evalua-tion on CPU and GPU)模型根据图像大小及缩放要求来评估其在CPU和GPU中重采样效率,据此选择基于CPU或者GPU的重采样方式,并且在此基础上建立了CPU-GPU图像块缓存策略。试验表明:根据RECG模型,对于不同的重采样比例,采取不同的重采样方式,有效地加速了重采样速度。更多还原

【Abstract】 In large-scale image visualization applications,an image pyramid based on a quadtree is built to support to read and show images of special scope and resolution quickly. However,when the special resolution is not saved in the multi-resolution pyramid,it is still necessary to obtain the image of the specified resolution by resampling an image from the pyramid. Generally,the resampling process is implemented on CPU or GPU. In this paper,a RECG model is proposed to evaluate the resampling efficiency of images on CPU or GPU,through that a resampling method on CPU or GPU is selected to speed up the resampling process. Moreover,a CPU and GPU cache strategy is created to minimize the number of times. Image data is read from the drive. Experiments show that,compared with general image resampling process,the RECG deal with different resolution image quickly by choosing fitted sampling method on CPU or GPU.更多还原