Systems and methods for content-aware enhancement of images
AssigneeShinyfields Limited
InventorsDmitry Sytnik, Irina Lisova, Andrey Frolov
Life of this patent
Estimating the term…
Abstract
Exemplary embodiments are directed to a system for content-aware enhancement of an image. The system includes an interface configured to receive as input an original image, and a processing device in communication with the interface. The processing device is configured to process the original image to detect one or more objects in the original image, generate an in-focus person mask of the original image for one or more in-focus people in the original image, and apply one or more enhancements to areas of the original image excluding the in-focus person mask.
Claims16
A system for content-aware enhancement of an image, the system comprising:
- an interface configured to receive as input an original image;
- and
- a processing device in communication with the interface, the processing device configured to:
- (i) process the original image using a neural network to detect one or more objects in the original image;
- (ii) generate an in-focus person mask of the original image for one or more in-focus people in the original image;
- and
- (iii) apply one or more enhancements to areas of the original image excluding the in-focus person mask,
- wherein the processing device generates a person mask encompassing one or more movie in the original image and an out-of-focus mask encompassing out-of-focus areas in the original image, and
- the processing device generates the in-focus person mask, the in-focus person mask being a difference between the person mask and the out-of-focus mask.
A system for content-aware enhancement of an image, the system comprising:
- an interface configured to receive as input an original image;
- and
- a processing device in communication with the interface, the processing device configured to:
- (i) process the original image using a neural network to detect one or more objects in the original image;
- (ii) generate an in-focus person mask of the original image for one or more in-focus people in the original image;
- and
- (iii) apply one or more enhancements to areas of the original image excluding the in-focus person mask,
- wherein the processing device generates a grayscale image based on the original image, and
- wherein the processing device generates a modulated image, the modulated image generated by application of a modified Laplacian effect on the grayscale image.
A method for content-aware enhancement of an image, the method comprising:
- receiving as input at an interface an original image;
- detecting one or more objects in the original image with a neural network;
- generating an in-focus person mask of the original image for one or more in-focus people in the original image;
- and
- applying one or more enhancements to areas of the original image excluding the in-focus person mask,
- generating a person mask encompassing one or more people in the original image and an out-of-focus mask encompassing out-of-focus areas in the original image, and
- wherein the in-focus person mask is a difference between the person mask and the out-of-focus mask.
A non-transitory computer-readable medium storing instructions at least for content-aware enhancement of an image that are executable by a processing device, wherein execution of the instructions by the processing device causes the processing device to:
- receive as input at an interface an original image;
- detect one or more objects in the original image with a neural network;
- generate an in-focus person mask of the original image for one or more in-focus people in the original image;
- and
- apply one or more enhancements to areas of the original image excluding the in-focus person mask,
- generate a person mask encompassing one or more people in the original image and an out-of-focus mask encompassing out-of-focus areas in the original image, and
- wherein the in-focus person mask is a difference between the person mask and the out-of-focus mask.
Dependencies are inferred from the claim text.