DisCO: Portrait Distortion Correction with Perspective-Aware 3D GANs

Portrait photos captured from a near-range distance often suffer from undesired perspective distortions. DisCO corrects these perspective distortions and synthesizes more pleasant views by virtually enlarging focal length and camera-to-subject distance.

More results on cropped faces

Correct in-the-wild distorted images collected by us and by Zhao+

Abstract

Close-up facial images captured at close distances often suffer from perspective distortion, resulting in exaggerated facial features and unnatural/unattractive appearances. We propose a simple yet effective method for correcting perspective distortions in a single close-up face. We first perform GAN inversion using a perspective-distorted input facial image by jointly optimizing the camera intrinsic/extrinsic parameters and face latent code. To address the ambiguity of joint optimization, we develop focal length reparametrization, optimization scheduling, and geometric regularization. Re-rendering the portrait at a proper focal length and camera distance effectively corrects these distortions and produces more natural-looking results. Our experiments show that our method compares favorably against previous approaches regarding visual quality. We showcase numerous examples validating the applicability of our method on portrait photos in the wild.

Method overview

Comparisons

Visual comparisons on our collected images

Input
Ours
#Fried+

Visual comparisons on images collected by Zhao+

Input
Ours
Fried+
Zhao+

Visual comparisons on CMDP

Input
Fried+
Ours
Reference

Quantitative comparisons

We conduct comparisons on CMDP. Results of Fried+ is borrowed from their demo page. #Fried+ denotes our re-implementation of Fried+.


Ablation study

Input
w/o all
w/o all +camera opt
w/o distance init
w/o scheduling
w/o reparameterization
Ours
Reference

Reference

  1. Yajie Zhao, Zeng Huang, Tianye Li, Weikai Chen, Chloe LeGendre, Xinglei Ren, Jun Xing, Ari Shapiro, and Hao Li, Learning Perspective Undistortion of Portraits, ICCV, 2019.
  2. Ohad Fried, Eli Shechtman, Dan B Goldman, and Adam Finkelstein, Perspective-aware Manipulation of Portrait Photos, ACM TOG (Proc. SIGGRAPH), 2016.
  3. Burgos-Artizzu, Xavier, Ronchi, Matteo Ruggero, & Perona, Pietro (2022). Caltech Multi-Distance Portraits (CMDP) (1.0) [Data set]. CaltechDATA. https://doi.org/10.22002/D1.20110

Acknowledgements

Special thanks to Yajie Zhao for providing their results and data; Ohad Fried for sharing their results on web.

Our collected in-the-wild images are from internet under common creative. Sources are here.

The website design is adapted from SunStage and HyperNeRF.