Gaussian filter-based dark channel prior for image dehazing enhancement

Abstrak

The presence of haze in an image is one of the challenges in computer vision tasks, such as remote sensing, object monitoring, and traffic monitoring applications. The hazy image is considered to contain noise and it can interfere with the image analysis process. Thus, image dehazing becomes a necessity as part of image enhancement. Dark channel prior (DCP) is one of the images dehazing methods that works based on a physical degradation model and utilizes low-intensity values from outdoor image characteristics. The DCP method generally consists of some steps, which are finding the dark channel and gradient image, estimating the sky region, atmospherical light, and transmission map, and reconstructing the dehazed image. This study introduces image dehazing by utilizing the Gaussian filter combined with the DCP method to increase the sharpness and accentuate the details of hazy images. Experimental results show that the proposed method could produce dehazed images with a visual quality is 18.94 dB on average or an increase of 11.91% compared to the original hazy image with a similarity index is 66.71% on average or an increase of 8.10%. Therefore, it is expected that this study can contribute to the image dehazing method enrichment based on DCP.