Double-buffer ping-bang parallel-structure image processing optimization method based on DMA (direct memory access)

Abstract

The invention relates to a double-buffer ping-bang parallel-structure image processing optimization method based on DMA (direct memory access). The method comprises the following steps: a DSP (digital signal processor) is externally connected with an SDRAM (synchronous dynamic random access memory), images to be processed are stored in the SDRAM, dividing the image data in the SDRAM into 2N image bands, opening up odd-even buffer areas for storing odd-even image band data in the DSP, then utilizing fast data transmission capability of the DSP and a parallel working mode of a DMA controller and a CPU (central processing unit) to realize optimization processing for the images stored in the SDRAM. In the double-buffer ping-bang parallel-structure image processing optimization method, the utilization ratio of the CPU and the DMA controller is effectively improved, the operation speed of image processing algorithm is accelerated, the performance of the DSP in the aspect of real-time image processing is improved, and the method can be widely applied in the fields of real-time image processing, such as scene matching, image fusion and target tracking and the like, and also can be used for the field of military.

Description

technical field [0001] The invention relates to an image processing optimization method, in particular to a DMA-based double-buffering ping-pong parallel structure image processing optimization method, which belongs to the field of embedded image processing algorithm optimization. Background technique [0002] With the development of digital processing methods, digital image processing technology has been more and more applied in scientific research, industrial production or management departments. Image processing techniques such as scene matching, image fusion, and target tracking are widely used in the aerospace field. Real-time image processing has the characteristics of large amount of data and high correlation of data processing, and the real-time application environment determines that the real-time system must have powerful computing capabilities. The development platform based on DSP can provide it with high-speed data processing capability, and its programmability...

AI Technical Summary

Problems solved by technology

[0003] The image processing algorithm has the characteristics of complex calculation, large amount of data, and high correlation of data processing. The complex operation means that the algorithm includes a large amount of data access operations. The large amount of image data makes it impossible to store all images in the internal cache of the DSP. Data processing High correlation leads to data migration and data processing must be carried out sequentially, these characteristics make the image processing algorithm on the DSP platform have shortcomings such as long calculation time and low real-time performance

At present, the introduction of relevant topics in the public literature mainly focuses on single professional directions such as digital image processing algorithms, DSP program optimization methods, and embedded system design schemes. There are few introductions to embedded real-time image processing algorithm optimization methods, and the optimization methods are different. , there is no discussion on the optimization algorithm design of double-buffered ping-pong parallel structure image processing based on DMA on the DSP platform

Images