A Novel Approach for Image Dehazing via Spatial and Channel Feature Fusion

The quality of images collected outdoors is compromised owing to hazy weather, resulting in contrast reduction and degradation. To solve this problem, a new dehazing algorithm that combines spatial and channel feature maps was proposed. The algorithm comprises of two parts: image dehazing and image enhancement. The first part uses the anti-symmetric nature of the Involution operator and the convolution operator to extract the spatial and channel feature maps of the hazy image completely. This part can restore the dehazed image by combining it with a deformed atmospheric scattering model. In the second part, a contrast-limited adaptive histogram equalization algorithm is used to improve the dehazed image. This process can improve the contrast and brightness of an image and enhance its detailed information. The experiment used the structural similarity index measure(SSIM), peak signal-to-noise ratio(PSNR), and other parameters to compare the algorithm with three other common dehazing algorithms to verify the dehazing effect and robustness of the algorithm. The experimental results show that the proposed algorithm improved the SSIM and PSNR by 4.72% and 15.96% respectively.