Systems and methods for image vignetting replacement
AssigneeAdobe Inc.
InventorsArjun Haarith G, Ankur Murarka
Life of this patent
Estimating the term…
Abstract
Systems and methods for image generation are provided. An aspect of the systems and methods includes obtaining a digital image including a content area and a border area that includes at least two corners of the digital image; generating a border mask for the border area based on a symmetry of the border area with respect to the at least two corners; and generating a generated image based on the border mask, wherein the generated image includes the content area and a generated area corresponding to the border area, and wherein the generated area depicts an extension to content from the content area.
Claims16
A method for image generation, comprising:
- obtaining a digital image including a content area and a border area;
- generating, using a mask component, a border mask for the border area by identifying a first preliminary mask segment and a second preliminary mask segment based on a color of the border area, flipping the first preliminary mask segment to match an orientation of the second preliminary mask segment, identifying a pixel of the first preliminary mask segment that is located outside of an intersection of the flipped first preliminary mask segment and the second preliminary mask segment, excluding the identified pixel from the first preliminary mask segment to obtain a border mask segment, and combining the border mask segment and the second preliminary mask segment to obtain the border mask;
- and
- generating, using an image generation network, a generated image based on the border mask, wherein the generated image includes the content area and a generated area corresponding to the border area, and wherein the generated area depicts an extension to content from the content area.
A non-transitory computer readable medium storing instructions that, when executed by a processor, cause the processor to:
- obtain a digital image including a content area and a border area including a first corner and a second corner of the digital image;
- generate, using a mask component, a border mask for the border area by identifying a first preliminary mask segment and a second preliminary mask segment, flipping the first preliminary mask segment to match an orientation of the second preliminary mask segment, identifying an intersection of the flipped first preliminary mask segment and the second preliminary mark segment, and flipping a copy of the intersection to match an orientation of the first preliminary mask segment, wherein the border mask comprises the intersection and the copy of the intersection;
- and
- generate, using an image generation network, a generated image based on the border mask, wherein the generated image includes the content area and a generated area corresponding to the border area.
A system for image generation, comprising:
- one or more processors;
- one or more memory components coupled with the one or more processors;
- a mask component designed to generate a border mask for a border area of a digital image by identifying a first preliminary mask segment and a second preliminary mask segment based on a color of the border area, flipping the first preliminary mask segment to match an orientation of the second preliminary mask segment, identifying a pixel of the first preliminary mask segment that is located outside of an intersection of the flipped first preliminary mask segment and the second preliminary mask segment, excluding the identified pixel from the first preliminary mask segment to obtain a border mask segment, and combining the border mask segment and the second preliminary mask segment to obtain the border mask;
- and
- an image generation network including parameters stored in the one or more memory components and trained to generate a generated image based on the border mask, wherein the generated image includes a content area of the digital image and a generated area corresponding to the border area, and wherein the generated area depicts an extension to content from the content area.
Dependencies are inferred from the claim text.