Sky images enhancement under complex weather conditions

An effective enhancement method based on the adaptive local variance is presented. Firstly, To handle the noise problem, a smooth method based on the heat kernel is proposed. To be precise, a weighted undirected graph is used to represent the image pixel lattice, such that the anisotropic diffusion across this weighted graph-structure can be captured by the heat equation. Secondly, by using the method, the image noise is reduced through the heat kernel which keeps the structural information of the image. To further reinforce the image clarity, a modified method based on the local variance is proposed, which may provide dynamic sub-blocks to strengthen the local pixel intensity. The experimental results show that this modified method extends the dynamic range of local image, provides the immunity to noise and may obtain well enhancement images and this modified method offers a favorable condition for next steps of processing.