Generating natural images with direct Patch Distributions Matching
This addresses the challenge of efficient and high-quality image generation for computer vision applications, offering a non-adversarial alternative that is incremental in improving upon traditional patch-based methods.
The paper tackles the problem of generating realistic images by directly matching patch distributions between generated and training images, using the Sliced Wasserstein Distance to achieve results often superior to single-image GANs with no training and generation in seconds.
Many traditional computer vision algorithms generate realistic images by requiring that each patch in the generated image be similar to a patch in a training image and vice versa. Recently, this classical approach has been replaced by adversarial training with a patch discriminator. The adversarial approach avoids the computational burden of finding nearest neighbors of patches but often requires very long training times and may fail to match the distribution of patches. In this paper we leverage the recently developed Sliced Wasserstein Distance and develop an algorithm that explicitly and efficiently minimizes the distance between patch distributions in two images. Our method is conceptually simple, requires no training and can be implemented in a few lines of codes. On a number of image generation tasks we show that our results are often superior to single-image-GANs, require no training, and can generate high quality images in a few seconds. Our implementation is available at https://github.com/ariel415el/GPDM