Image frame processing optimizing method based on Davinci platform

Abstract

The invention provides an image frame processing optimizing method based on a Davinci platform. The image frame processing optimizing method comprises the following steps in sequence: designating three video image frame memory addresses in an internal memory at the same time, wherein the three video image frame memory addresses are the destination addresses of EDMA (Enhanced Direct Memory Access); identifying the three video image frequency memory addresses into buffer 1, buffer 2 and buffer 3; sequentially storing the image data frames into the plurality of video image memory addresses by virtue of the EDMA according to the sequence circulation of buffer 1, buffer 2 and buffer 3; and writing the image data frames into some video image frame memory addresses by the EDMA, and processing the image data of two frames in another two video image frame memory addresses by a CPU (Central Processing Unit) at the same time, thus obtaining a pingpong mechanism under the execution of the CPU. By adopting the image frame processing optimizing method based on the Davinci platform, the utilization rate of the CPU is increased, the images of two frames are read while processing the image frame by the CPU, thus the step of executing memcpy is saved, and as a result, the executing efficiency is greatly improved.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an image frame processing optimization method based on a Da Vinci platform with higher efficiency without memcpy operation. Background technique [0002] At present, the demand and application of embedded intelligent transportation products are becoming more and more extensive. Whether it is a high-speed checkpoint or an intersection, vehicle detection equipment is required. Instead of the traditional ground induction coil, pure video vehicle detection has been more and more applied. An important method of pure video vehicle detection is the grayscale image frame difference algorithm based on virtual coils. [0003] The frame difference algorithm is to subtract two or more frames of continuous images pixel by pixel to remove static or slow-moving objects and backgrounds. It is the most direct way to eliminate static objects in two consecutive images and provide moving o...

AI Technical Summary

Problems solved by technology

The frame difference algorithm in step 4 can be optimized with some algorithm strategies and inline assembly, but the memcpy operation in step 5 is time-consuming and cannot be optimized. The larger the image is, the more time-consuming it will be, and it cannot even be completed within 40ms. If it is required to be separated by 2 frames or more If the frame is poor, the number of memcpy will be more

Images