Edge-Preserving Texture Transfer Using the Laplacian Pyramid

Texture plays an important role in our perception of photographic images. It can be used to differentiate objects, draw the eye to salient regions, and convey information about a subject. However, post processing photographs in image editing software can be time consuming, often requiring advanced skills to achieve convincing effects. We introduce a multi-scale method, based on the Laplacian pyramid, to introduce auxiliary textures into photographic images. The final result retains the structural characteristics from the input, including edges, colour, and high-contrast existing texture, while enhancing the image with fine-scale details. In addition, we extend patch-based texture synthesis to include a guidance channel so that texture structures are aligned with an orientation field, obtained through the image structure tensor. Both the original structure tensor calculation and its subsequent smoothing are aided with geodesic knowledge so that our orientation fields are both edge preserving and smooth.