Image Smoothing via Local Structure Variation

To address the imbalance smoothing problem in structure and texture, we propose a local structure variation-based edge-preserving image smoothing method. Firstly, the structure measure and texture measure are built using variations in the local statistics of pixel values at distinct locations. Secondly, measures are converted into regularization items. A variation smoothing model is established, and multi-smoothness images are obtained by sparse solution. Finally, a structure extraction and fusion model is developed to maintain structure and suppress texture in complicated texture images by adjusting the smoothness of different image regions. The high-quality smoothing performance of our method is confirmed by the subjective and objective experiments.