Estimation Algorithm of Atmospheric Light based on Ant Colony Optimization
Wenbo Zhang, Xiaorong Hou · 2017
In the haze removal algorithm based on the atmospheric scattering model, atmospheric light value is an important parameter and its accuracy directly influences the quality of the haze removal results. Existing haze removal methods adopt a fixed number of atmospheric light candidate points for the clustering and statistic estimation of atmospheric light, with the maximum point cluster including candidate points. However, due to small candidate point samples, there is a high amount of error in the estimation of atmospheric light in a statistical sense. In order to solve that problem, this study uses the approach of dividing threshold values to select candidate points of atmospheric light and ant colony optimization to cluster point-clusters of atmospheric light. In this way, the number of candidate point samples and therefore the accuracy of estimate results of atmospheric light are increased. In addition, in order to improve the computational efficiency of the algorithm, this study first uses the K-means algorithm for a preliminary aggregation of atmospheric light candidate points, and then utilizes ant colony optimization to improve the clustering results. The experimental results prove that atmospheric light obtained using this method leads to more natural haze removal results, and further improves the index of image quality evaluation in the haze removal results.