Parameterized Brushstroke Style Transfer

TL;DR

A style transfer method that represents the image in the brush stroke domain instead of the RGB domain, which has better visual improvement over pixel-based methods is discussed.

Abstract

Computer Vision-based Style Transfer techniques have been used for many years to represent artistic style. However, most contemporary methods have been restricted to the pixel domain; in other words, the style transfer approach has been modifying the image pixels to incorporate artistic style. However, real artistic work is made of brush strokes with different colors on a canvas. Pixel-based approaches are unnatural for representing these images. Hence, this paper discusses a style transfer method that represents the image in the brush stroke domain instead of the RGB domain, which has better visual improvement over pixel-based methods.

Figures (4)

• Figure 1: B represents the output of Gatys' style transfer method, while C is the result of our proposed approach. Visually, our method produces an image that more closely resembles artistic brush strokes compared to Gatys'. D is the output after pixel optimization, where the brush strokes are blended, resulting in a more cohesive and refined representation.
• Figure 2: The top row shows the results of Gatys' style transfer approach, while the bottom row displays the output of the proposed method. In our approach, parameterized brush strokes are passed through a differentiable renderer, which maps them onto the canvas. The content and style losses are then calculated, and gradients are backpropagated through the renderer to optimize the brush strokes. The image is taken from kotovenko2021rethinkingstyletransferpixels
• Figure 3: Zoomed view of brush strokes with pixel optimization.
• Figure 4: Style transfer applied to a human image.