AlphaGAN: Generative adversarial networks for natural image matting

TL;DR

This paper introduces AlphaGAN, a novel GAN-based approach for natural image matting that improves spatial detail preservation and achieves state-of-the-art results, especially on fine structures like hair.

Contribution

AlphaGAN is the first GAN designed for natural image matting, utilizing dilated convolutions and adversarial training to enhance spatial detail and visual quality.

Findings

Achieves state-of-the-art gradient error on alphamatting benchmark.

Excels at fine structures like hair in image matting.

Comparable results on other benchmark metrics.

Abstract

We present the first generative adversarial network (GAN) for natural image matting. Our novel generator network is trained to predict visually appealing alphas with the addition of the adversarial loss from the discriminator that is trained to classify well-composited images. Further, we improve existing encoder-decoder architectures to better deal with the spatial localization issues inherited in convolutional neural networks (CNN) by using dilated convolutions to capture global context information without downscaling feature maps and losing spatial information. We present state-of-the-art results on the alphamatting online benchmark for the gradient error and give comparable results in others. Our method is particularly well suited for fine structures like hair, which is of great importance in practical matting applications, e.g. in film/TV production.