Image dehazing aims to restore image clarity and visual quality by reducingatmospheric scattering and absorption effects. While deep learning has madesignificant strides in this area, more and more methods are constrained bynetwork depth. Consequently, lots of approaches have adopted parallel branchingstrategies. however, they often prioritize aspects such as resolution,receptive field, or frequency domain segmentation without dynamicallypartitioning branches based on the distribution of input features. Inspired bydynamic filtering, we propose using cascaded dynamic filters to create amulti-branch network by dynamically generating filter kernels based on featuremap distribution. To better handle branch features, we propose a residualmultiscale block (RMB), combining different receptive fields. Furthermore, wealso introduce a dynamic convolution-based local fusion method to mergefeatures from adjacent branches. Experiments on RESIDE, Haze4K, and O-Hazedatasets validate our method's effectiveness, with our model achieving a PSNRof 43.21dB on the RESIDE-Indoor dataset. The code is available athttps://github.com/dauing/CasDyF-Net.