Open CL accelerated video defogging method

Abstract

The invention discloses an open CL accelerated video defogging method. The method adopts an OpenCL to defog a foggy video, and specifically comprises the following steps: initializing a FFmpeg multimedia framework; initializing an OpenCL architecture; optimizing a FHRUSI defogging algorithm through the adoption of the OpenCL to realize parallel video defogging processing; expanding the parallel video defogging processing to the FFmpeg multimedia framework and running a parallel video defogging processing sub-program. The OpenCL parallel programming method and GPU parallel architecture are used for processing the parallel acceleration of the video defogging processing and the real-time defogging of a high-definition video, the complexity of the defogging algorithm is lowered, and the parallelism of the algorithm is improved, the execution time of the defogging algorithm is greatly reduced, and the high-speed and high-definition defogging effect is achieved.

Description

technical field [0001] The invention relates to the technical field of computer image processing, in particular to a video defogging method. Background technique [0002] Smog is a common weather phenomenon. When the smog is severe, the visibility and visibility of the road will deteriorate, and the road and environmental information obtained by the driver will be insufficient, which is very likely to cause vicious traffic accidents. Therefore, how to increase the amount of information about the surrounding environment that drivers can obtain under low-visibility weather conditions has always been a research hotspot in the field of intelligent transportation. [0003] In the field of intelligent transportation, the existing auxiliary equipment and algorithms have good adaptability to night, but the effect on image defogging is poor. Low visibility is due to insufficient light, and the feedback information of objects can be obtained through infrared equipment; and low visibi...

AI Technical Summary

Problems solved by technology

The method of image enhancement does not require a physical model. It mainly highlights the details of the image by enhancing the contrast. The algorithm is relatively simple to implement, but the dehazing effect is not as good as the image restoration algorithm.

Most of the algorithms based on image restoration have good defogging effects, but the complexity of the algorithm is very high, the amount of calculation is large, and the implementation process takes a long time

[0005] At present, the research focus of the dehazing algorithm based on a single image is mostly focused on the dehazing algorithm and the image’s dehazing effect, and the research on algorithm acceleration is still very limited. Occupies a lot of memory, requires a long running time, cannot bring ideal results for project realization, and is difficult to be widely used

In addition, in terms of video defogging processing, since the resolution of high-definition video is much higher than that of ordinary video, the calculation amount of real-time defogging of high-definition video is even greater, and the defogging speed is slower

The dark channel hypothesis is such a defogging algorithm. Using this algorithm to process a 600×400 picture on the CPU takes about 10-20s, which obviously cannot meet the real-time requirements of many projects. Therefore, shorten the dark channel Assuming that the running time of the dehazing algorithm is necessary

Images