Mesh Denoising via Genetic Algorithm

Mesh denoising is an essential process in many geometric applications. We describe a simple and efficient mesh denoising approach based on genetic algorithm. The raw mesh is smoothed using a floating-point genetic algorithm that is more flexible than the usual binary genetic algorithms, and can handle non-smooth regions containing several local extrema. The fitness function selected is a weighted linear combination of the triangle aspect ratio and the Laplacian distance at each node of the triangular mesh. Compared with widely-used gradient descent based schemes, our method avoids specifying the iteration step size, and performs better at challenging regions with rich geometric features. Extensive qualitative and quantitative experiments demonstrate that our approach can effectively remove noise from meshes.