Real-time gas image enhancement algorithm based on guided filter simplification kernel

Aiming at the problem of low signal-to-noise ratio and weak imaging in infrared imaging of toxic and harmful gases, this paper proposes an algorithm for real-time enhancement of gas images based on the gain mask of the simplified kernel of the guided filter to improve the visual effects of gas infrared images. First, based on the gas movement between the two frames, the difference information between the two frames is amplified. Then, the differentially amplified image is decomposed into a detailed layer containing gas texture details and noise information and a base layer containing gas contour parts through a guided filter. The kernel of the guide filter is simplified to identify edges, textures and flat areas in the detail layer. A gain mask based on the guide filter simplification kernel is designed for the detail layer, which does not filter out the gas texture information and retains the edge information of the image while suppressing noise. Then the enhanced detail layer and the base layer are recombined to obtain an enhanced image. Finally, a segmentation algorithm based on the Gaussian mixture model is performed on the enhanced image to extract the gas area, and an adaptive connected domain is designed to remove misjudged front sights, so as to separate the gas area from the background. The experimental results show that the gas enhancement algorithm proposed in this paper can segment the gas area to a large extent, has strong robustness to the environment and exhibits good realtime performance.