| CPC G06T 7/337 (2017.01) [G06T 3/147 (2024.01); G06T 5/77 (2024.01); G06T 7/40 (2013.01); G06T 7/50 (2017.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30201 (2013.01)] | 24 Claims |
|
1. A method comprising:
aligning landmark points between multiple distorted images to generate multiple aligned images, wherein the multiple distorted images exhibit perspective distortion in at least one face appearing in the multiple distorted images, and wherein the landmark points are aligned to correct baseline disparities between the multiple distorted images caused by differences in at least one of: sensor sensitivities, calibrations, focal lengths, and baseline distance between imaging sensors;
predicting a depth map using a disparity estimation neural network that receives the multiple aligned images as input;
generating a warp field using a selected one of the multiple aligned images;
performing a two-dimensional (2D) image projection on the selected aligned image using the depth map and the warp field to generate an undistorted image, wherein the undistorted image includes one or more missing pixels as a result of the 2D image projection; and
filling in the one or more missing pixels in the undistorted image using an inpainting neural network to generate a final undistorted image.
|