A Effective Method Using Wavelet Transform for Haze and Noisy Image

Abstract

In order to enhance the visualization of low-quality images and adapt to the practical work of public security picture detection and visual detection in haze weather, we integrate the currently widely applicable processing methods and adopt a new processing method. This method uses the quad-tree method to solve the problem of the selection defect of the priori atmospheric light value of the dark channel. Aiming at the interference effect between the wavelet frequency division reconstruction coefficients, the adaptive correlation method under the gradient is applied. For noise that gains along with wavelet coefficients, we gives a noise reduction method by compromises the threshold. The experimental results show that the hierarchical contrast of the image is fine, the color fidelity is fine, and the indicators are generally better than other algorithms.