Systems and methods for extracting patches from digital images
A system including one or more processors and one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to perform operations: analyzing a respective foreground of one or more related digital images to identify a respective skin region in the respective foreground of the one or more related digital images; when the respective skin region is not identified, analyzing the respective foreground of the one or more related digital images to identify a respective head region in the respective foreground; extracting one or more patches from respective non-head region remainders of the respective foreground; and coordinating displaying a patch of the one or more patches on an electronic display based on one or more scores for the one or more patches. Other embodiments are described herein.
1 . A system comprising:
one or more processors; and
one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to perform operations comprising:
analyzing a respective foreground of one or more related digital images to identify a respective skin region in the respective foreground of the one or more related digital images;
when the respective skin region is not identified, analyzing the respective foreground of the one or more related digital images to identify a respective head region in the respective foreground of the one or more related digital images;
extracting one or more patches from respective non-head region remainders of the respective foreground of the one or more related digital images; and
coordinating displaying a patch of the one or more patches on an electronic display based on one or more scores for the one or more patches.
2 . The system of claim 1 , wherein the one or more scores for the one or more patches indicate one or more probabilities of the one or more patches being representative of an item in the one or more related digital images.
3 . The system of claim 1 , wherein the operations further comprise:
when the respective head region is identified, clustering a non-head remainder of the respective foreground of the one or more related digital images into one or more clusters; and
determining the one or more scores for the one or more patches of the one or more related digital images.
4 . The system of claim 1 , wherein extracting the one or more patches further comprises:
creating one or more matrices from the one or more related digital images; and
extracting the one or more patches from the respective foreground of the one or more related digital images using the one or more matrices.
5 . The system of claim 4 , wherein each matrix of the one or more matrices comprises:
respective matrix dimensions matching image dimensions of a corresponding digital image of the one or more related digital images; and
a respective entry for each respective pixel in the corresponding digital image of the one or more related digital images.
6 . The system of claim 5 , wherein the respective entry for each respective pixel identifies each respective pixel as at least one of:
a background pixel when the respective pixel is not in the foreground;
a skin pixel when the respective pixel is in the respective skin region; or
a clustered pixel when the respective pixel is grouped into one or more clusters, wherein, when the respective head region is identified, the one or more clusters are formed by clustering a non-head remainder of the respective foreground of the one or more related digital images.
7 . The system of claim 1 , wherein the operations further comprise:
before analyzing the respective foreground of the one or more related digital images, resizing the one or more related digital images to a common size.
8 . The system of claim 1 , wherein the operations further comprise:
before analyzing the respective foreground of the one or more related digital images, identifying, using an image segmentation algorithm, the respective foreground of the one or more related digital images.
9 . The system of claim 1 , wherein analyzing the respective foreground of the one or more related digital images further comprises:
determining, by one or more colorimetric properties, the respective skin region in the one or more related digital images; and
identifying pixels within a predetermined color range labeled as the respective skin region.
10 . The system of claim 1 , wherein the operations further comprise:
when the respective skin region is identified, clustering, using a Gaussian mixture model, a non-skin remainder of the respective foreground of the one or more related digital images into one or more clusters.
11 . A method implemented via execution of computing instructions configured to run at one or more processors and configured to be stored at non-transitory computer-readable media, the method comprising:
analyzing a respective foreground of one or more related digital images to identify a respective skin region in the respective foreground of the one or more related digital images;
when the respective skin region is not identified, analyzing the respective foreground of the one or more related digital images to identify a respective head region in the respective foreground of the one or more related digital images;
extracting one or more patches from respective non-head region remainders of the respective foreground of the one or more related digital images; and
coordinating displaying a patch of the one or more patches on an electronic display based on one or more scores for the one or more patches.
12 . The method of claim 11 , wherein the one or more scores for the one or more patches indicate one or more probabilities of the one or more patches being representative of an item in the one or more related digital images.
13 . The method of claim 11 further comprising:
when the respective head region is identified, clustering a non-head remainder of the respective foreground of the one or more related digital images into one or more clusters; and
determining the one or more scores for the one or more patches of the one or more related digital images.
14 . The method of claim 11 , wherein extracting the one or more patches further comprises:
creating one or more matrices from the one or more related digital images; and
extracting the one or more patches from the respective foreground of the one or more related digital images using the one or more matrices.
15 . The method of claim 14 , wherein each matrix of the one or more matrices comprises:
respective matrix dimensions matching image dimensions of a corresponding digital image of the one or more related digital images; and
a respective entry for each respective pixel in the corresponding digital image of the one or more related digital images.
16 . The method of claim 15 , wherein the respective entry for each respective pixel identifies each respective pixel as at least one of:
a background pixel when the respective pixel is not in the foreground;
a skin pixel when the respective pixel is in the respective skin region; or
a clustered pixel when the respective pixel is grouped into one or more clusters, wherein, when the respective head region is identified, the one or more clusters are formed by clustering a non-head remainder of the respective foreground of the one or more related digital images.
17 . The method of claim 11 further comprising:
before analyzing the respective foreground of the one or more related digital images, resizing the one or more related digital images to a common size.
18 . The method of claim 11 further comprising:
before analyzing the respective foreground of the one or more related digital images, identifying, using an image segmentation algorithm, the respective foreground of the one or more related digital images.
19 . The method of claim 11 , wherein analyzing the respective foreground of the one or more related digital images further comprises:
determining, by one or more colorimetric properties, the respective skin region in the one or more related digital images; and
identifying pixels within a predetermined color range labeled as the respective skin region.
20 . The method of claim 11 further comprising:
when the respective skin region is identified, clustering, using a Gaussian mixture model, a non-skin remainder of the respective foreground of the one or more related digital images into one or more clusters.