Image scaling hardware acceleration method, device and system and readable storage medium

Abstract

The invention discloses an image scaling hardware acceleration method and device. Th method comprises steps of storing position information of data needing to be extracted in a position table according to an image scaling ratio of a proximity interpolation method; extracting the data in the original picture according to the position information in the position table through a position look-up table method; storing data extracted from the original picture into a barrel-shaped shift register; and outputting the data from the foremost end of the barrel-shaped shift register according to the output bandwidth. According to the embodiment of the invention, parallel pipeline operation of data input and output is realized by adopting the position look-up table method and the barrel-shaped shifteralgorithm. Tests show that the speed of the algorithm is remarkably increased compared with the serial output speed of a common algorithm. The processing speed of the image data is prevented from being limited by DDR bandwidth, accumulation stagnation during data processing is relieved, and the whole deep learning efficiency is improved.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an image scaling hardware acceleration method, device, system and readable storage medium. Background technique [0002] In computer image processing and computer graphics, image scaling is the process of resizing a digital image. Image scaling is a non-trivial process that requires a trade-off between processing efficiency and the smoothness and clarity of the results. Among various image scaling algorithms, proximity interpolation is the most basic and simplest image scaling algorithm. Image scaling is an indispensable step in deep learning algorithms. During data processing, images that need to be scaled are usually cached in DDR memory, and the processing speed of image data is limited by the bandwidth of DDR, resulting in data processing stagnation. Ordinary algorithms will inevitably lead to longer time-consuming image scaling, which will affect the efficiency of ...

AI Technical Summary

Problems solved by technology

Image scaling is an indispensable step in deep learning algorithms. During data processing, images that need to be scaled are usually cached in DDR memory, and the processing speed of image data is limited by the bandwidth of DDR, resulting in data processing stagnation. Ordinary algorithms will inevitably lead to longer time-consuming image scaling, which will affect the efficiency of the entire deep learning

Images