Image stylization is a task that involves transforming an input image into a new image that has a different style, while preserving the content of the original image. The goal of image stylization is to create visually appealing images with a specific style or aesthetic, such as impressionism, cubism, or surrealism. It can also be used to make images more visually appealing for specific applications, such as social media or advertising.
Instance Normalization (also known as contrast normalization) is a normalization layer where: $$ y_{tijk} = \frac{x_{tijk} - \mu_{ti}}{\sqrt{\sigma_{ti}^2 + \epsilon}}, \quad \mu_{ti} = \frac{1}{HW}\sum_{l=1}^W \sum_{m=1}^H x_{tilm}, \quad \sigma_{ti}^2 = \frac{1}{HW}\sum_{l=1}^W \sum_{m=1}^H (x_{tilm} - \mu_{ti})^2. $$ This prevents instance-specific mean and covariance shift simplifying the learning process. Intuitively, the normalization process allows to remove instance-specific contrast information from the content image in a task like image stylization, which simplifies generation.