Noise removal using an adaptive Euler's elastica-based model

We consider the Euler's elastica-based model for an image noise removal problem. Several recent works have demonstrated that Euler's elastica-based model performs better than the celebrated total variation model on preserving image features in smooth regions during denoising. On the other hand, an adaptive weighting scheme on total variation is a technique for well-restoring local features of an image. Inspired by these two strategies, we propose an adaptive Euler's elastica-based model to handle both local features of image and image features in smooth regions simultaneously. Numerically, we apply the alternating direction method of multipliers to solve this non-smooth and non-convex model. Experimental results on both natural and synthetic images illustrate the efficiency of the proposed method.