作者: Moatasem Chehaiber
DOI:
关键词: Scale-invariant feature transform 、 Gaussian pyramid 、 Pixel 、 Computer vision 、 Kernel (image processing) 、 Artificial intelligence 、 Mathematics
摘要: The present invention relates to the parallel calculation of convoluted data. In particular, Gaussian pyramid construction and processing image data, such as repeatedly data for use in a SIFT algorithm. This is achieved by providing method obtaining plurality difference images from an original defined pixels, said comprising: Providing blurring convolution functions, each functions increasing degree upon image; establishing Dif, calculating between two different degrees image, convolving with obtain images.