Pencil style transfer of digital illustrations using LIC and depth estimation

In recent years, with the widespread use of social media, the demand for image processing technology has increased, and style transfer has attracted attention. In particular, conversion to pencil-style drawings, which are impressive and easily attract the viewer's attention, is also expected to be a technology that can efficiently generate images suitable for use in products and advertisements. However, most previous studies have assumed photographic images as input and have not sufficiently addressed cases in which digital illustrations are used instead. Thus, there is a problem that the characteristic exaggerated expressions, such as line drawings with strong and weak lines, seen in digital illustrations, tend to be overly emphasized. To address this limitation, we propose a method that combines simple image processing techniques with deep learning to achieve pencil-style transfers of digital illustrations while minimizing the loss of visual impression.