ebook img

Method and apparatus for classifying and identifying images PDF

41 Pages·2013·2.86 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Method and apparatus for classifying and identifying images

US006549660B1 United States Patent (12) (10) Patent N0.: US 6,549,660 B1 Lipson et a1. (45) Date of Patent: Apr. 15, 2003 (54) METHOD AND APPARATUS FOR OTHER PUBLICATIONS CLASSIFYING AND IDENTIFYING IMAGES _ Mahesh Venkatraman et al., Zero Crossmgs of a Non—Or (75) Inventors; Pamela R_ Lipson, Cambridge, M A thogonal Wavkelet Transform for Object Location, IEEE, p. (US); W. Eric L. Grimson, Lexington, 57_60> 1995 M A (Us); Pawan Sinha, Cambridge, W. Niblack et al., The QBIC Project: Quering Images by M A (Us); Tomaso poggio, weuesley, Content Using Color, Texture, and Shape SPIE vol. 1908, p. MA (Us) 173—168, 1993* Rohini K Srihari, “Automatic Indexing & Content—based (73) Assignee: Massachusetts Institute of Retrieval of Captioned Images”, IEEE, pp. 49—56, 1995 .* Technology, Cambridge, MA (US) Kathleen PereZ—LopeZ, “Comparison of Subband Features for Automatic Indexing of Scienti?c Image Database”, ( * ) Notice: Subject‘ to any disclaimer, the term of this SPIE, V01- 2185, PP~ 94—94, 1994* Patent 15 extended or adlusted under 35 Akio Yamamoto, “Extraction of Object Features & its Appli U-S-C- 154(k)) by 0 days- cation to Image Retrieval”, The Transactions of the IEICE, vol. E72, No. 6, pp. 771—781, 1989.* (21) Appl- NOJ 09/270,995 Hong Jiang Zhang et al., “A Scheme for Visual Features (22) Filed: Man 17’ 1999 based Image Indexing”, SPIE vol. 2420, pp. 36—46, 1995.* Related US. Application Data (List Continued on next page) _ _ _ _ Primary Examiner—Samir Ahmed (63) i320niglélgnggsggg’pggatgog6gags/600378’ ?led on Feb‘ (74) Attorney, Agent, or Firm—Daly, Crowley & Mofford, (51) Int: . Cl.7a ............' ......'. ...a. .....a. ......' ............. .. G06K 9/62 (57) ABSTRACT An image processing system for generating a class model (52) US. Cl. ...................................... .. 382/224; 345/581 3.0m an Image by ldennfymg Ida?“ relanenshlns between ifferent properties of different image regions, mcludes a _ region partitioner a relationship processor, a template gen (58) Fleld of seagcshz 330852/ erator and an image detector. The image processing system / 5_ ’ ’_ 5 ’ 5 ’ 55’ 5 ’ 5 ’ utilizes a class model de?ned by one or more relative 164’ 22 346864436151 34’545 15872’ relationships betWeen a plurality of image patches. The ’ ’ ’ ’ / _ relative relationships describe the overall organization of (56) References Cited images Within an image class. The relative relationships are encoded in a global deformable template Which can be used US. PATENT DOCUMENTS to classify or detect images. The class model may be pre-de?ned or generated by the image processing system. In 4,633,506 A 12/1986 Kato ......................... .. 382/56 one embodiment, the class model is generated from a low 4,742,558 A * 5/1988 Ishibashi et a1. 382/240 resolution image. 5,157,743 A 10/1992 Maeda et a1. . . . . . . . . . . . .. 382/56 5,164,992 A 11/1992 Turk et a1. . . . . . . . . . . . . .. 382/2 29 Claims, 21 Drawing Sheets 5,384,868 A 1/1995 Maeda et a1. . . . . . . . . . . . .. 382/56 5,390,259 A 2/1995 Withgott et a1. ......... .. 382/9 5,404,435 A 4/1995 Rosenbaum ....... .. 395/147 5,440,401 A 8/1995 Parulski et a1. ........... .. 358/342 (14 of 21 Drawing Sheet(s) Filed in Color) IMAGE DATABASE [12 IMAGE RETRIEVAL IMAGE STORAGE! DEVICE SCANNING DEVICE 26 DEFORMABLE TEMPLATE TEMPLATE IMAGE GENERATOR IMAGE TRANSFER DETECTOR APPARATUS /_ 24 2a {- 22 SUBSAMPLER H REGION RELATIONSHIP Ag?gg?gg" PARTITIONER PROCESSOR US 6,549,660 B1 Page 2 OTHER PUBLICATIONS “Feature Identi?cation As An Aid to Content—based Image Retrieval”, by Ramesh et al., published in SPIE, vol. 2420, Transactions of the Institute of Electronics, Information and pp. 2—11. Communication Engineers of Japan, vol. E72, No. 6, Jun. “Flexible Montage Retrieval for Image Data”, by Sakamoto 1989, pp. 771—781, XP000050590 Yamamoto et al: “Extrac et al., published in SPIE vol. 2185, pp. 25—33. tion of Object Features and Its Application To Image “Image Retrieval based on Multidimensional Feature Prop Retrieval”. erties”, by Ang, et al., published in SPIE vol. 2420, pp. Proceedings of the Region 10 Annual International Confer 47—57. pp. 84—94. ence Tenco, Singapore, Aug. 22—26, 1994, vol. 1, Aug. 22, “Image Retrieval Using Context Vectors: First Results”, by 1994, Chan T K Y, pp. 407—411, XP000529510. Gallant et al., published in SPIE vol. 2420, pp. 82—94. “An Image Database System With Fast Image Indexing “Image Searching in a Shipping Product”, by Will EquitZ, Capability Based On Color Histograms”. published in SPIE vol. 2420, pp. 186—196. Proceedings of the International Conference on Image Pro “0—Index of Imagebase”, by Ying Liu, published in SPIE cessing (ICIP), Washington, Oct. 23—26, 1995, vol. 1, Insti vol. 2420, pp. 68—81. tute of Electrical and Electronics Engineers, pp. 342—345, “Interactive Indexing into Image Databases”, by Michael J. XP000624245, Lin H—C et al: “Color Image Retrieval Based SWain, published in SPIE vol. 1908, pp. 95—103. on Hidden Markov Models”. “Interpretation of Natural Scenes Using Multi—Parameter Record of the Asilomar Conference on Signals, Systems and Default Models and Qualitative Constraints”, by Hild et al., Computers, Paci?c Grove, Oct. 30—Nov. 2, 1994, vol. 2, No. published in IEEE, 1993, PP. 497—501. Conf. 28, Institute of Electical and Electronics Engineers, “Methodology for the Representation, Indexing and pp. 1341—1345, XP000533864 Sclaroff S et al: “Search By Retrieval of Images by Content”, by Petrakis et al., pub Shape Examples: Modeling Nonrigid Deformation”. lished in Image and Vision Computing vol. 11, No. 8, Oct. “Perceiving and RecogniZing Three—Dimensional Forms”, 1993, pp. 504—521, and a last page unnumbered. Ph.D. Thesis, Chapter 7, Dept. of EECS, MIT, pp. 1—19, and “Modeling Complex Objects in Content—Based Image Figs. 1—18, Sep. 1995. Retrieval”, by Youssef Lahlou, published in SPIE vol. 2420, “A Content—Based Indexing Technique Using Relative pp. 104—115. Geometry Features”, Hou et al.; published in SPIE vol. “Photobook: Content—Based Manipulation of Image Data 1662, 1992, pp. 59—68. bases”, by Pentland, et al., published in SPIE vol. 2185, Feb. “A Hierarchical Approach to Feature Indexing”, by Grosky 6—10, 1994, pp. 1—24. pp. 84—94. et al., published in SPIE vol. 1662 Image Storage and “Photobook: Tools for Content—Based Manipulation of Retrieval Systems 1992, pp. 9—20. Image Databases”, Pentland et al., published in SPIE vol. “A Scheme for Visual Feature Based Image Indexing”, by 2185, pp. 34—47. Zhang et al., published in SPIE vol. 2420, pp. 36—46. Query by Image and Video Content: The QBIC System, by “A 2—D String Matching Algorithm for Conceptual Pictorial Flickner et al., published in IEEE, 1995, pp. 23—32. Queries”, by Chang et al., published in SPIE vol. 1662 “Query by Image Content and Its Applications”, Ashley et Image Storage and Retrieval Systems (1992), pp. 47—58. al., published in Computer Science/Mathematics, Mar. 17, “Adaptive Filtering and Indexing for Image Databases”, by 1995, pp. 1—7 With ?rst tWo pages Unnumbered. Alexandrov et al., published in SPIE vol. 2420, pp. 12—23. “Retrieval of Similar Pictures on Pictorial Databases”, by “Automatic Indexing and Content—Based Retrieval of Cap Chang et al., published in Pattern Recognition, vol. 24, No. tioned Images”, by Srihari, published in IEEE, 1995, pp. 7, 1991, pp. 675—680. 49—56. “Query by Image Example: The Candid Approach”, by “Automatic and Semi—Automatic Methods for Image Anno Kelly et al., published in SPIE vol. 2420, pp. 238—248. tation and Retrieval in QBIC”, by Ashley et al., published in “Searching Images Using Ultimedia Manager”, Treat et al., SPIE vol. 2420, pp. 24—35. published in SPIE vol. 2420, pp. 204—213. pp. 84—94. “Chabot: Retrieval from a Relational Database of Images”, “Shape Analysis for Image Retrieval”, by Domenico Tegolo, by Ogle et al., published in Computer, pp. 40—48. published in SPIE vol. 2185, pp. 59—69. “Comparison of Subband Features for Automatic Indexing Shape Similarity—Based Retrieval in Image Database Sys of Scienti?c Image Databases”, by PereZ—LopeZ et al., tems, by Gary et al., published in SPIE vol. 1662 Image published in SPIE vol. 2185, pp. 84—94. “Complex Queries for a Query—By—Content Image Data Storage and Retrieval (1992). pp. 2—8. base”, by Lee et al., published in Computer Science, Nov. “Similar—Shape Retrieval in Shape Data Management”, Mehrotra et al., published in IEEE, Sep. 1995, pp. 57—62. 21, 1994, pp. 1—22. “Content—Based Image Retrieval Systems”, by Gudivada et “Similarity of Color Images”, by Stricker et al., published in al., published in IEEE, 1995, pp. 18—21, last page unnum SPIE vol. 2420, pp. 381—392. bered. “Similarity Measures for Image Databases”, by Jain et al., “Ef?cient Color Histogram Indexing for Quadratic Form published in SPIE vol. 2420, pp. 58—65. Distance Functions”, Hafner et al., published in IEEE, “Similarity Searching in Large Image Databases”, by Transactions on Pattern Analysis and Machine Intelligence, Petrakis et al., published in CS—TR—3388, Dec. 1994, pp. vol. 17 No. 7, Jul. 1995, pp. 729—736. 1—33, With cover page unnumbered. “Evaluation of an Application—Independent Image Informa “Snakes: Active Contour Models”, Kass et al., published in tion System”, by Oakley et al., published in SPIE vol. 2185, International Journal of Computer Vision, 1988, pp. pp. 107—118. 321—331 With last page unnumbered. “Feature Extraction from Faces Using Deformable Tem “The QBIC Project: Querying Images By Content Using plates”, by Yuille et al., published in IEEE, 1989, pp. Color, Texture, and Shape”, by Niblack et al., published in 104—109, including a last page not numbered. SPIE vol. 1908, 1993, pp. 173—187. US 6,549,660 B1 Page 3 “Visual Image Retrieval for Applications in Art and Art “Object Recognition via Image Invariants”,by PaWan Sinha, History”, by Holt et al., published in SPIE vol. 2185, pp. available on—line at http://WWW.aimit.edu/projects/cbcl/We 70—81. b—homepage/eWb—homepage.html. Individual pages each numbered 1, (total of 9 pages) by Virage, Inc., available on—line at: http://WWW.virage.com/. * cited by examiner U.S. Patent Apr. 15,2003 Sheet 2 0f 21 US 6,549,660 B1 U.S. Patent Apr. 15,2003 Sheet 3 0f 21 US 6,549,660 B1 I START I 36 \ PROVIDE REFERENCE IMAGE 37 \" IDENTIFY IMAGE PROPERTIES 38 EXTARCT LOW FREQUENCY IMAGE COMPONENTS? Yes I 40 \ GENERATE LOW FREQUENCY IMAGE 42 \ SUBSAMPLE REFERENCE IMAGE 43 \ PARTITION IMAGE INTO IMAGE REGIONS V 44 \ COMPUTE REGION RELATIONSHIPS FIG. 2 U.S. Patent Apr. 15,2003 Sheet 4 0f 21 US 6,549,660 B1 46 ASSIGN PATCH COLOR 1/ MORE PATCHES‘): 50 REDUCE REIATIQNS'TO EQUIVALENCE CLASSES l 52 STORE RELATIVE - RELATIONSHIPS FIG. 2A U.S. Patent Apr. 15,2003 Sheet 5 0f 21 US 6,549,660 B1 IDENTIFY CONSISTENT 53 ATTRIBUTES RELATIONSHIPS / AT EACH PATCH ACROSS ALL REFERENCE IMAGES . N0 54 YES 55 ASSDCIATE ATFR‘BUTES . BETVV‘EEN REGfONS’ GENERATE GOMPOSITE /“ 56 IMAGE T l 51 SH THRESHOLD 1/ l 58 GENERATE A DEFORMABLE TEMPLATE 59 MORE NO TO 36 (FIG. ) REFERENCE IMAGES U.S. Patent Apr. 15, 2003 Sheet 6 0f 21 US 6,549,660 B1 U.S. Patent Apr. 15,2003 Sheet 7 0f 21 US 6,549,660 B1

Description:
cation to Image Retrieval”, The Transactions of the IEICE, vol. E72, No. An image processing system for generating a class model. (52) US. Cl Record of the Asilomar Conference on Signals, Systems and. Computers .. Sheet 16 0f 21. US 6,549,660 B1. 33%}. 2.3%. @361. 213%” 2% gig'? 'Egg
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.