Image processing
Out of all the things that image generative models can do, image processing is probably the one that produces the best results for commercial use. In our context, image processing refers to applying some transformation to existing images to produce new ones. We will look at the three applications of image processing in this section – image inpainting, image compression, and image super-resolution (ISR).
Image inpainting
Image inpainting is the process of filling in missing pixels of an image so that the result is visually realistic. It has practical applications in image editing, such as restoring a damaged image or removing obstructing objects. In the following example, you can see how image inpainting is used to remove people in the background. We first fill the people in with white pixels, then we use a generative model to fill in the pixels:
Figure 10.2 – Image inpainting using DeepFillv2 to remove people in the background...
