SYSTEMS AND METHODS OF MULTICOLOR SEARCH OF IMAGES
Systems and methods are provided for receiving at least a first query color, and searching an electronic catalog including a plurality of product images for the first query color to determine a similarity measure between the first query color and a product image of a plurality of product images. The similarity measure may be determined by determining a Euclidean distance between values in a three-dimensional color space for the first query color and a target color of the product image, and determining the similarity measure between the query color and the product image by determining a sum of the similarity measures from all target colors on the product image, weighted by the coverage of each target color. The search results may be transmitted based on the searching of the electronic catalog including the plurality of product images for the first query color.
1 . A method comprising:
performing, at a server, object detection and background subtraction for each product image of a plurality of product images of an electronic catalog stored in at least one storage device communicatively coupled to the server;
applying, at the server, a color network over each product image to determine one or more feature vectors;
determining, at the server, the nearest neighbor product in the one or more feature vectors, for every point in a color space;
searching, at the server, the plurality of the product images of the electronic catalog for a fixed color or a plurality of colors; and
transmitting, via a communications network coupled to the server, search results based on the searching of the electronic catalog for the fixed color or the plurality of colors.
2 . The method of claim 1 , wherein features of the one or more feature vectors include at least one selected from the group consisting of: dominant colors, color ratios, and masks.
3 . The method of claim 1 , where the determining the nearest neighbor comprises:
performing locality sensitive hashing (LSH) of a Hamming distance between the one or more feature vectors.
4 . The method of claim 3 , wherein the LSH is performed using an approximate principal direction tree.
5 . The method of claim 1 , wherein the searching at the server for the fixed color or the plurality of colors is a weighted search.
6 . A system comprising:
at least one storage device to store a plurality of product images; and
a server, communicatively coupled to the at least one storage device, to perform object detection and background subtraction for each product image of the plurality of product images of an electronic catalog, to apply a color network over each product image to determine one or more feature vectors, to determine the nearest neighbor product in the one or more feature vectors, for every point in a color space, to search the plurality of the product images of the electronic catalog for a fixed color or a plurality of colors, and to transmit, via a communications network coupled to the server, search results based on the searching of the electronic catalog for the fixed color or the plurality of colors.
7 . The system of claim 6 , wherein features of the one or more feature vectors include at least one selected from the group consisting of: dominant colors, color ratios, and masks.
8 . The system of claim 6 , where the server determines the nearest neighbor by performing locality sensitive hashing (LSH) of a Hamming distance between the one or more feature vectors.
9 . The system of claim 8 , wherein the LSH is performed by the server using an approximate principal direction tree.
10 . The system of claim 6 , wherein the searching at the server for the fixed color or the plurality of colors is a weighted search.