Parallel image processing method and system

Abstract

The invention provides a parallel image processing method. The parallel image processing method comprises the steps of collecting image data at a first moment; dispatching the image data collected atthe first moment and transmitting the image data to a standby first processing thread; the processing thread calls algorithms in sequence according to the arrangement sequence of the algorithms to process the image data collected at the first moment; judging whether the image data acquired at the first moment is processed or not at the nth moment; if the image acquired at the first moment is not processed completely, continuing to acquire image data; scheduling the image data collected at the nth moment and transmitting the image data to the nth standby processing thread; the processing threadcalls algorithms in sequence according to the arrangement sequence of the algorithms to process the image data collected at the nth moment; and if the image acquired at the first moment is processed,stopping acquiring the image data, and displaying the image.

Description

technical field [0001] The invention provides a parallel image processing method and system, in particular a parallel image processing method and system capable of utilizing spare computer resources. Background technique [0002] In the process of image processing, the collected raw image data is generally processed to eliminate noise, improve contrast, or extract key signals to meet specific applications in different industries. Each processing process can be called an algorithm, and generally multiple algorithms are executed sequentially to generate a final image for display. [0003] Assume that the time consumption of algorithm A is TA, the time consumption of algorithm B is TB, the time consumption of algorithm C is TC, the time points of collection are t1, t2, t3, etc., and the average time interval of collecting an image is T, then the system The acquisition frame rate is 1 / T. Then, if the final image is to be displayed at the speed of image acquisition, the total t...

AI Technical Summary

Problems solved by technology

However, not all algorithms can be executed in parallel internally. These algorithms can only use one CPU core at runtime, and process all pixels in a traversal manner.

In addition, in some algorithms, each pixel needs to be averaged with several pixels around the pixel, and it is easy to form a discontinuous boundary at the boundary when the block is calculated in parallel.

In order to solve the discontinuity phenomenon, the block image needs to overlap several widths or heights, which means that the calculation is repeated at the boundary, and then the overlapping part is discarded when merging to eliminate or reduce the boundary phenomenon, which brings the calculation Expansion of resources and complexity of final synthesis

Images