Image recognition methods
DCFirst Claim
1. A method of converting an image into a compact form for comparisons and other purposes, comprising the steps of:
- a) providing an image;
b) associating a grid of points with the image, each point having an initial position;
c) transforming the set using one or more attributes of the image by warping the grid such that some or all of the points move from their initial positions to different positions, thereby generating a unique key representative of the image.
6 Assignments
Litigations
0 Petitions
Accused Products
Abstract
Physical objects, including still and moving images, are transformed into more compact forms for identification and other purposes using a method unrelated to existing image-matching systems which rely on feature extraction. An auxiliary construct, preferably a warp grid, is associated with an image, and a series of transformations are imposed to generate a unique visual key for identification, comparisons, and other operations. Search methods are also disclosed for matching an unknown image to one previously represented in a visual key database. Broadly, a preferred search method sequentially examines candidate database images for their closeness of match in a sequential order determined by their a priori match probability. Thus, the most likely match candidate is examined first, the next most likely second, and so forth. With respect to the recognition of video sequences and other information streams, inventive holotropic stream recognition principles are deployed, wherein the statistics of the spatial distribution of warp grid points is used to generate index keys. The invention is applicable to various fields of endeavor, including governmental, scientific, industrial, commercial, and recreational object identification and information retrieval.
321 Citations
29 Claims
-
1. A method of converting an image into a compact form for comparisons and other purposes, comprising the steps of:
-
a) providing an image;
b) associating a grid of points with the image, each point having an initial position;
c) transforming the set using one or more attributes of the image by warping the grid such that some or all of the points move from their initial positions to different positions, thereby generating a unique key representative of the image. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22)
providing a second image;
performing steps b) and c) on the second image to generate a second key; and
comparing the second key with the key in the database to determine if the images are similar.
-
-
14. The method of claim 1, further including the steps of:
-
performing steps b) and c) on a plurality of images to generate a key for each;
storing the keys in a database;
providing a search image;
performing steps b) and c) on the search image to generate a search key; and
comparing the search key to the keys stored in the database to determine if the search image is among the plurality.
-
-
15. The method of claim 14, further including the steps of:
sequentially ordering the keys in the database in accordance with a predetermined match probability prior to the step of comparing the search key.
-
16. The method of claim 1, including the step of deriving the image from a book, periodical or other printed matter.
-
17. The method of claim 16, wherein the image is derived using a camera, scanner, or other apparatus operative to convert an image into electronic form.
-
18. The method of claim 13, including the step of deriving the search image from another database, software application, hyperlink or web page.
-
19. The method of claim 1, further including the steps of:
-
providing a plurality of images in the form of sequential frames portraying motion; and
analyzing the statistics of the construct transformations associated with the images to generate the keys.
-
-
20. The method of claim 16, further including the step of:
-
providing a search image; and
determining if the search image is contained in one of the frames using a dynamic decision tree that changes in accordance with the statistical likelihood of a match.
-
-
21. The method of claim 20, wherein the frames, search image, or both are derived from a video or other form of motion picture.
-
22. The method of claim 21, wherein the image may not be reconstructed from the key.
-
23. A method of associating with an image a measurement vector, comprising the steps of:
-
a) superimposing a lattice of grid points over the image;
b) associating with each grid point in the lattice a subset of grid points that constitute its connectivity pattern;
c) sampling the image at each lattice grid point to obtain a local measurement of the image;
d) adapting the positions of each of the lattice grid points in the image based upon the positions and sampled measurements of each grid point in its connectivity pattern; and
e) iteratively repeating steps (c) and (d) until lattice grid point positions achieve equilibrium and remain essentially unchanged in position with further iterations or iteratively repeating steps (c) and (d) a fixed number of times sufficient to achieve near equilibrium of lattice grid points and (f) preparing a measurement vector of the displacements of all the iteratively adapted lattice grid points from step (d) from their initial positions. - View Dependent Claims (24, 25, 26, 27, 28, 29)
-
Specification