中文摘要：

双线性插值算法在数字图像处理中有广泛的应用，但计算速度慢．为提高其计算速度，提出一种基于图形处理器加速的双线性插值并行算法．主要利用Wallis变换双线性插值中各分块之间的独立性适合GPU并行处理架构的特点，把传统串行双线性插值算法映射到CUDA并行编程模型，并从线程分配，内存使用，硬件资源划分等方面进行优化，来充分利用GPU的巨大运算能力．实验结果表明，随着图像分辨率的增大，双线性内插并行算法可以把计算速度提高28倍．

英文摘要：

Bilinear interpolation is widely used to process digital image, but is time-consuming. In order to speed up, a bilinear interpolation parallel algorithm based on Graphic Processing Unit（ GPU） acceleration is proposed. According to the independence among different bilinear interpolation data blocks in Wallis transform suitable for the parallel architecture of GPU, we mapped the traditional bilinear interpolation algorithm to CUDA parallel programming model with optimizing thread allocation, memory usage, hardware resources partition to take full advantage of the huge arithmetic capability of GPU. The experimental result shows that bilinear interpolation parallel algorithm could get a more than 28 x speedup as the image resolution increases.