Image real-time dehazing method based on graphics processor

Abstract

The invention discloses a real-time dehazing method for images based on a graphics processor, which mainly solves the shortcomings in the prior art of using a computer central processing unit (CPU) which has a slow processing speed and is not suitable for real-time dehazing of high-resolution images. The implementation steps are: 1. Initialize the computer, input the hazy image and read the image data; 2. Calculate the dark channel map of the input fog image; 3. Calculate the atmospheric light value based on the dark channel map; 4. Calculate the input fog image Initial transmittance map; 5. Perform guided filtering on the initial transmittance map to obtain the final transmittance map; 6. Restore the scene based on the final transmittance map and output a clear and haze-free image. The invention can ensure good defogging effect and has fast processing speed, meets the real-time defogging requirements of 1080p images, and can be used for real-time defogging processing of high-resolution images.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a real-time image defogging method, which can be used in the fields of outdoor target recognition and tracking, road visual monitoring, satellite remote sensing monitoring and the like. Background technique [0002] The generation of fog image is mainly due to the scattering effect of turbid media such as fog and water vapor in the atmosphere, which seriously attenuates the light intensity received by the camera. Therefore, in order to ensure the clarity of the image and the recognizability of information, and to correct the color shift of the image due to atmospheric light, the image defogging algorithm is usually used to process the fog image. The image defogging algorithm based on the dark channel prior physically defoggs the image, and the effect is obvious and real and natural, so it is widely used in the field of image defogging. [0003] However, due to the...

AI Technical Summary

Problems solved by technology

The disadvantage of this method is that the parallelism of the whole implementation scheme is not enough, some processes still use the CPU for serial processing, the GPU calculation part does not fully exploit the reusability of data, and does not use high-speed memory such as texture storage and shared storage to access data The mode is optimized so that it can only process low-resolution video images in real time. When the image resolution is slightly higher, it cannot achieve fast processing. For example, for a video image with a resolution of 720×576, its processing speed is only 21 frame / second

Obviously, it is difficult to achieve real-time processing for high-definition video images such as 1080p

Images