Method and apparatus for comparing objects in images
AssigneeCanon Kabushiki Kaisha
InventorGetian Ye
Life of this patent
Estimating the term…
Abstract
A method of comparing objects in images. A dictionary determined from a plurality of feature vectors formed from a test image and codes formed by applying the dictionary to the feature vectors is received, the dictionary being based on a difference in mean values between the codes. Comparison codes are determined for the objects in the images by applying the dictionary to feature vectors of the objects in the images. The objects in the images are compared based on the comparison codes of the objects.
Claims17
A method to compare objects in images from a target domain, the method comprising:
- receiving a dictionary determined from a plurality of feature vectors and codes, wherein the plurality of feature vectors was formed from a test image in the target domain, the codes were formed by applying the dictionary to the plurality of feature vectors of the target domain, and the dictionary is based on source domain training data and on a difference in mean values between the formed codes;
- determining comparison codes for the objects in the images by applying the dictionary to feature vectors of the objects in the images;
- and
- comparing the objects in the images based on the comparison codes of the objects.
An apparatus to compare objects in images from a target domain, the apparatus comprising:
- a receiving unit configure to receive a dictionary determined from a plurality of feature vectors and codes, wherein the plurality of feature vectors was formed from a test image in the target domain, the codes were formed by applying the dictionary to the plurality of feature vectors of the target domain, and the dictionary is based on source domain training data and on a difference in mean values between the formed codes;
- a determining unit configure to determine comparison codes for the objects in the images by applying the dictionary to feature vectors of the objects in the images;
- and
- a comparing unit configure to compare the objects in the images based on the comparison codes of the objects.
A non-transitory computer-readable medium storing a program to cause a computer to execute a method for an apparatus to compare objects in images from a target domain, the method comprising:
- receiving a dictionary determined from a plurality of feature vectors and codes, wherein the plurality of feature vectors was formed from a test image in the target domain, the codes were formed by applying the dictionary to the plurality of feature vectors of the target domain, and the dictionary is based on source domain training data and on a difference in mean values between the formed codes;
- determining comparison codes for the objects in the images by applying the dictionary to feature vectors of the objects in the images;
- and
- comparing the objects in the images based on the comparison codes of the objects.
A system for comparing objects in images, the system comprising:
- a memory for storing data and a computer program;
- and
- a processor coupled to the memory for executing the program, the program comprising instructions for:
- receiving a dictionary determined from a plurality of feature vectors and codes, wherein the plurality of feature vectors was formed from a test image in the target domain, the codes were formed by applying the dictionary to the plurality of feature vectors of the target domain, and the dictionary is based on source domain training data and on a difference in mean values between the formed codes,
- determining comparison codes for the objects in the images by applying the dictionary to feature vectors of the objects in the image, and
- comparing the objects in the images based on the comparison codes of the objects.
A method for a computer system to compare gallery objects in a first image captured by a first camera and a query object in a second image captured by a second camera, wherein the first camera and the second camera are controlled by the computer system and the gallery objects, and the query object include feature vectors, the method comprising:
- receiving, in the computer system, a dictionary determined in a cloud server as follows:
- collecting target domain training data from the first camera and the second camera during a training phase of operation in advance of capturing the gallery objects and the query object, wherein the collected target domain training data includes training images of objects of a same type as the gallery objects and the query object and was collected without knowledge of the identities of the gallery objects and the query object,
- acquiring pre-exiting source domain training data from sources other than the first camera and the second camera, and
- learning the dictionary on the cloud server based on the target domain training data and the source domain training data;
- determining comparison codes for the gallery objects, and the query object by applying the received dictionary to the feature vectors of the objects in the images;
- and
- matching objects between the first image and the second image by comparing the comparison codes determined using the received dictionary.
Dependencies are inferred from the claim text.