Guided Image Filter Inspired Improved Single Image Dehazing Method

EasyChair Preprint no. 9655

Abstract

With the recent proliferation of industries, traffic and wildfires, the haze is becoming a common atmospheric phenomenon which has a critical impact on the computer-vision based acquisition systems. To address the image degradation due to the hazy weather, we propose an improved physical model-based single image dehazing method. The proposed dehazing process includes the calculation of the atmospheric light value, estimation of the transmission map and finally, the restoration of scene radiance from the hazed image. The main contribution of this work is to combine the recent advances in the atmospheric light value calculation techniques with the utilization of refined guided filter by employing a colour image as guidance for the estimation of an optimized transmission map. The detailed qualitative and quantitative analysis of our proposed method with current state-of-the-art techniques, reveal that dehazed images generated from our algorithm are of improved visual quality with sufficient image details.

Keyphrases: Atmospheric light scattering model, Guided image filter, Haze, Koschmeider’s law, single image dehazing

@Booklet{EasyChair:9655,
  author = {Muhammad Shan Saleem and Gulistan Raja and Junaid Mir},
  title = {Guided Image Filter Inspired Improved Single Image Dehazing Method},
  howpublished = {EasyChair Preprint no. 9655},

  year = {EasyChair, 2023}}