Image dehazing using non-local haze-lines and multi-exposure fusion

Images captured under haze conditions suffer from color distortion and low saturation due to the light propagates through scattering particles, causing light intensity attenuation and direction deflection, which affects the imaging quality of the visual system. To deal with these issues, we propose an image dehazing method based on non-local haze-line and multi-exposure fusion, called NHMF. Specifically, we first equalize the brightness and color of the input image according to a multi-scale fusion strategy. Meanwhile, we use the dark channel prior based on a local window to solve the atmospheric light value. Afterward, we employ the preprocessed image to obtain fog lines with better generalization performance that enhances the estimation of the transmission rate. Furthermore, we introduce weighted least-squares filtering to refine transmittance estimation accuracy further and ultimately employ an atmospheric scattering model to reverse process the haze-free image. Our extensive experiments on three image enhancement datasets demonstrate the effectiveness of our approach in quantitative and qualitative dehazing of images with haze. Moreover, our method exhibits excellent generalization performance in dehazing remote sensing images and enhancing underwater images.