Region refocusing for data-driven object localization
First Claim
1. A method for segmenting an image comprising:
- computing a first image signature for an input image;
identifying at least one similar image from a first set of images, based on the first image signature and image signatures of images in the first set of images;
generating a cropped region of the input image based on the identified at least one similar image;
computing a second image signature for the cropped region;
identifying at least one similar image from a second set of images, based on the second image signature and image signatures of images in the second set of images; and
segmenting the input image based on a segmentation map of at least one of the at least one similar image identified in the second set of images.
4 Assignments
0 Petitions
Accused Products
Abstract
A system and method are provided for segmenting an image. The method includes computing an image signature for an input image. One or more similar images are identified from a first set of images, based on the image signature of the input image and image signatures of images in the first set of images. The similar image or images are used to define a cropped region of the input image and a second image signature is computed, this time for the cropped region. One or more similar images are identified from a second set of images, based on the cropped image signature and the image signatures of images in the second set of images. The input image is segmented based on a segmentation map of at least one of the similar images identified in the second set of images.
-
Citations
25 Claims
-
1. A method for segmenting an image comprising:
-
computing a first image signature for an input image; identifying at least one similar image from a first set of images, based on the first image signature and image signatures of images in the first set of images; generating a cropped region of the input image based on the identified at least one similar image; computing a second image signature for the cropped region; identifying at least one similar image from a second set of images, based on the second image signature and image signatures of images in the second set of images; and segmenting the input image based on a segmentation map of at least one of the at least one similar image identified in the second set of images. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21)
-
-
22. A system for image segmentation comprising:
-
memory which stores; a signature generator configured for generating; a first image signature for an input image to be segmented, and a second image signature for a cropped region of the input image; a retrieval component configured for identifying at least one first similar image from an associated first set of images, based on the first image signature and image signatures of images in the first set of images; a segmentation component configured for identifying a segmented region of the input image based on segmentation maps of the identified at least one similar image from the first set; and a cropping component configured for cropping the input image to generate the cropped region, based on the identified segmented region; the retrieval component being configured for identifying at least one second similar image from an associated second set of images, based on the second image signature and image signatures of images in the second set of images; the segmentation component being configured for segmenting the input image based on segmentation maps of the identified at least one second similar image from the second set; and a processor which implements the signature generator, retrieval component, cropping component, and segmentation component. - View Dependent Claims (23)
-
-
24. A method for forming a system for image segmentation comprising:
-
providing a signature generator for generating a first image signature for an input image to be segmented and a second image signature for a cropped region of the input image; learning a first projection matrix, for projecting the first image signature into a first feature space, using image signatures of images drawn from a first training set and their respective segmentation maps; learning a second projection matrix, for projecting the second image signature into a second feature space, using image signatures of images drawn from a second training set and their respective segmentation maps; providing a retrieval component for; identifying at least one similar first image from a first set of images, similarity being computed based on the projected image signature of the input image and the projected image signatures of the images in the first set of images, the projected image signature of the input image being generated by projecting the input image'"'"'s image signature with the first projection matrix, and identifying at least one similar second image from a second set of images, similarity being computed based on the projected image signature of the cropped region and the projected image signatures of the images in the second set of images, the projected image signature of the cropped region being generated by projecting the cropped region'"'"'s image signature with the second projection matrix, the cropped region being generated based on the segmentation map of the identified at least one similar first image; and providing a segmentation component for segmenting the input image based on a segmentation map of at least one of the at least one similar second image.
-
-
25. A method for image segmentation comprising:
-
with a computer processor, cropping an input image to generate a cropped region of the input image, based on a computed similarity between a representation of the input image and representations of images in a first set of images for which respective segmentation maps are defined; identifying a target region within the cropped region, based on a computed similarity between a representation of the cropped region and representations of images in a second set of images for which respective segmentation maps are defined, the images in the second set being each derived from a respective image in the first set by cropping.
-
Specification