Making Images from Images: Interleaving Denoising and Transformation
This work addresses the challenge of creative image transformation for artists and designers, offering a novel approach that is incremental over prior optical illusion generation methods.
The paper tackles the problem of generating new images by rearranging regions of a source image, such as the Mona Lisa, into any desired subject matter, achieving this through a method that learns both image content and parameterized transformations simultaneously, with results improving as the number of regions increases.
Simply by rearranging the regions of an image, we can create a new image of any subject matter. The definition of regions is user definable, ranging from regularly and irregularly-shaped blocks, concentric rings, or even individual pixels. Our method extends and improves recent work in the generation of optical illusions by simultaneously learning not only the content of the images, but also the parameterized transformations required to transform the desired images into each other. By learning the image transforms, we allow any source image to be pre-specified; any existing image (e.g. the Mona Lisa) can be transformed to a novel subject. We formulate this process as a constrained optimization problem and address it through interleaving the steps of image diffusion with an energy minimization step. Unlike previous methods, increasing the number of regions actually makes the problem easier and improves results. We demonstrate our approach in both pixel and latent spaces. Creative extensions, such as using infinite copies of the source image and employing multiple source images, are also given.