Remote Sensing Image Dehazing Based on an Attention Convolutional Neural Network

Haze may affect the quality of optical remote sensing images, thus limiting scope their application. Remote image dehazing has become important in preprocessing, promoting use data and precision target recognition. Existing methods based on simplified atmospheric degradation models are not suitable for removal heterogeneous haze that exist images. For this purpose, study proposes an end-to-end convolutional neural network attention mechanism, which residual block structure combines both channel spatial mechanisms, establishes a synthetic high-resolution dataset full training. Thus, it obtains desired model. Finally, investigates model using GF-1 compares with existing methods. The results show proposed method improved similarity, color authenticity, residue level.