WO2008013679A1

WO2008013679A1 - Method and apparatus for automatically annotating images

Info

Publication number: WO2008013679A1
Application number: PCT/US2007/015887
Authority: WO
Inventors: Jay N. Yagnik
Original assignee: Google Inc.
Priority date: 2006-07-24
Filing date: 2007-07-11
Publication date: 2008-01-31
Also published as: CN101467145A; EP2044532A1; US8065313B2; US20080021928A1; CN101467145B

Abstract

One embodiment of the present invention provides a system that automatically annotates an image. During operation, the system receives the image. Next, the system extracts image features from the image. The system then identifies other images which have similar image features. The system next obtains text associated with the other images, and identifies intersecting keywords in the obtained text. Finally, the system annotates the image with the intersecting keywords.

Description

I

METHOD AND APPARATUS FOR AUTOMATICALLY

ANNOTATING IMAGES

Inventor: Jay N. Yagnik

BACKGROUND

Field of the Invention

[0001] The present invention relates techniques for automatically annotating images. More specifically, the present invention relates a technique that automatically annotates an image by searching for similar images and correlating text surrounding these similar images.

Related Art

[0002] The recent proliferation of high-bandwidth Internet connections presently makes it possible for millions of users to efficiently view images on the Internet. These developments have led to a tremendous increase in the number of images that are incorporated into web pages, web portals, and other web-based applications. Unfortunately, many images do not come with accompanying textual information, such as labels, captions, or titles, to describe the content of the images. This makes it extremely difficult to search for specific images because most existing search techniques are text based (e.g., a keyword search). Hence, it is highly desirable to annotate such images with relevant text, for example, by adding a set of keywords or a caption to describe the semantic content of an image.

[0003] Traditional techniques for annotating images are primarily manual, which can require human indexers to select keywords for thousands or, in some cases, millions of images. Hence, manual image annotation can be an extremely labor-intensive and expensive procedure. [0004] Other techniques have been developed to automatically annotate an image (see

"Formulating Semantic Image Annotation as a Supervised Learning Problem," G. Carneiro and N. Vasconcelos, Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, San Diego, 2005). These automatic image-annotation techniques can significantly reduce or eliminate the human effort required to annotate very large image collections.

[0005] However, these existing "automatic annotation" techniques typically still involve steps that require some degree of human assistance, such as requiring manual labeling of a set of ground truth data, or by requiring other types of human interaction or feedback. Unfortunately, this human assistance cannot be efficiently scaled to match the exponentially growing number of images which need to be annotated.

[0006] Hence, what is need is a method and an apparatus for automatically annotating an image without the above-described problems.

SUMMARY

[0007] One embodiment of the present invention provides a system that automatically annotates an image. During operation, the system receives the image. Next, the system extracts image features from the image. The system then identifies other images which have similar image features. The system next obtains text associated with the other images, and identifies intersecting keywords in the obtained text. Finally, the system annotates the image with the intersecting keywords.

[0008] In a variation on this embodiment, the system extracts the image features from the image by: (1) partitioning the image into tiles; and (2) extracting the image features from the tiles. [0009] In a further variation on this embodiment, for each given tile size in a set of differing tile sizes, the system partitions the image into tiles of the given tile size.

[0010] In a variation on this embodiment, prior to identifying the other images, the system matches the extracted image features against known image features in an image-feature library. [0011] In a further variation on this embodiment, the system combines the matched image features to form one or more image-feature-combinations for the image.

[0012] In a further variation, the system identifies the other images which have similar image features by identifying similar image-feature-combinations in the other images. [0013] In a variation on this embodiment, the system can extract the image features from the image by: (1) generating color histograms; (2) generating orientation histograms; (3) using a direct cosine transform (DCT) technique; (4) using a principal component analysis (PCA) technique; or (5) using a Gabor wavelet technique.

[0014] In a variation on this embodiment, the image features can be defined in terms of: (1) shapes; (2) colors; and (3) textures.

[0015] In a variation on this embodiment, the system identifies the other images by searching through images on the Internet.

[0016] In a variation on this embodiment, the system identifies the other images which have similar image features by using probability models. [0017] In a variation on this embodiment, prior to identifying the intersecting keywords, the system expands keywords in the obtained text by adding synonyms for the keywords.

[0018] In a variation on this embodiment, the system receives the image from a video.

BRIEF DESCRIPTION OF THE FIGURES

[0019] FIG. 1 presents a flowchart illustrating the process of automatically annotating an image in accordance with an embodiment of the present invention.

[0020] FIG. 2 illustrates a computer-based system for automatically annotating an image in accordance with an embodiment of the present invention.

DETAILED DESCRIPTION

[0021] The following description is presented to enable any person skilled in the art to make and use the invention, and is provided in the context of a particular application and its requirements. Various modifications to the disclosed embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments and applications without departing from the spirit and scope of the present invention. Thus, the present invention is not limited to the embodiments shown, but is to be accorded the widest scope consistent with the claims.

[0022] The data structures and code described in this detailed description are typically stored on a computer-readable storage medium, which may be any device or medium that can store code and/or data for use by a computer system. This includes, but is not limited to, volatile memory, non-volatile memory, magnetic and optical storage devices such as disk drives, magnetic tape, CDs (compact discs), DVDs (digital versatile discs or digital video discs), or other media capable of storing computer readable media now known or later developed.

Overview

[0023] Most image search and retrieval systems are text-driven wherein a user inputs keywords as queries, and a computer system, or more specifically, a search engine, processes the queries by searching through test associated with images in an image database. The search engine returns "relevant" images which are associated with textual information that matches the keywords. In this type of scheme, it is not possible for an un-annotated image to be retrieved, even if the image content is highly relevant to the keywords. Hence, providing relevant textual information for un-annotated images can facilitate search and retrieval operations for the images.

[0024] One embodiment of the present invention provides a technique that automatically annotates an image by generating text relevant to the image by using the vast image base that is available on the web. More specifically, one embodiment of the present invention extracts image features from a given image, and then searches through images on the Internet to identify a set of images which contain similar images features. Next, the text surrounding these images is obtained. The technique subsequently identifies common keywords in the obtained text which are used to annotate the given image. Note that this technique requires no human intervention, and utilizes many existing image-processing techniques to perform steps such as image feature extraction and identifying similar images on the web. Automatic Annotation Procedure

[0025] FIG. 1 presents a flowchart illustrating the process of automatically annotating an image in accordance with an embodiment of the present invention. [0026] The process starts with receiving an image, and a request to provide relevant textual information to the image (step 100). Note that the image may be formatted and stored in any digital format, including but are not limited to: JPEG, GIF, BMP, TIFF, PDF, PS, EMF, MNG, PNG, PSD, SWF and WMF. In addition, the image can also be obtained from still frames of a digital video. For example, we can extract an image from a single frame in a MPEG video and subsequently annotate this image.

[0027] Next, the process partitions the image into tiles of a specific size (step 102), and subsequently extracts image features from the tiles (step 104). Note that typical image features which can be extracted from these tiles can include, but are not limited to: shapes, colors, and textures. For example, several color features for a given tile in the image can be extracted, wherein the color features can include, but are not limited to: color histograms, a color histogram layout, and color moments. Furthermore, a number of image-processing techniques can be used to extract specific image features, wherein these image-processing techniques can include, but are not limited to: a direct cosine transform (DCT) technique, a principal component analysis (PCA) technique, and a Gabor wavelet technique. In one embodiment of the present invention, the image features can also include orientation histograms.

[0028] Also note that the present invention is not limited to features which are associated with tiles. The present invention can generally be used with any type of image feature, and is not limited to image features associated with tiles. For example, the present invention can be used with image features which are not associated with tile boundaries, such as attributes of a color histogram for the entire image.

[0029] The process then matches the extracted image features against known image features in an image-feature library (step 106). In one embodiment of the present invention, the known image features in the library were previously obtained using a similar image extraction technique. For example, the known image features can be represented by DCT coefficients if a DCT technique is used to extract the image features in step 104. In one embodiment of the present invention, matching the extracted image features against the known image features involves matching each extracted image feature to the nearest known image features in the library. [0030] Next, the process combines the matched known-image-features to form image-feature combinations for the image (step 108). More specifically, the process generates a set of image-feature combinations for the image partitioned at the specific tile sizes.

[0031] In one embodiment of the present invention, steps 102 to 108 are repeated iteratively for different tile sizes, wherein the tile sizes can range from only a number of pixels per tile up to a single tile containing the entire image. Specifically, for each tile size, the process partitions the image into tiles of that tile size and then extracts image features from the corresponding tiles. The process then matches extracted image features to the known image features and subsequently uses the known image features to produce a set of image-feature combinations for the image. Hence, the resulting image-feature combinations from step 108 comprise image-feature combinations obtained for different image tile sizes. Note that using different tile sizes facilitates capturing image features on multiple scales. Instead of matching to known image features, the system can alternatively restrict the values of the image features to fall into a limited set of discrete values. [0032] The process next identifies multiple other images which have similar image features and/or image-feature combinations (step 110). Specifically, the process searches through a large number of images from various sources. These image sources can include, but are not limited to, images stored in organized image databases, images embedded in webpages on the Internet, images associated with videos on the Internet, or any image that is linked to a web address on the Internet. Note that the difference between images on the

Internet and the images in the organized image databases is that the images on the Internet are typically not classified and are typically uncorrelated.

[0033] Determining if a particular image includes similar image features to the original image involves processing the particular image in the same manner as the original image to generate image features for the particular image. [0034] While comparing the image features and/or image-feature combinations in the original image with another image, various probability models can be applied. For example, these probability models can include anything from a simple counting technique, which counts the number of similar or identical features, to a complex "multivariate distribution technique," which assumes a multivariate normal distribution of features when comparing two images. (For a discussion of multivariate normal distributions, please see "http://en.wikipedia.org/wiki/Multivariate_gaussian_distribution".)

[0035] After the multiple similar images are identified, the process obtains text surrounding these images (step 112). In one embodiment of the present invention, the text is obtained around the image tag in the HTML file that includes the link to the image. In another embodiment, the text directly surrounding the actual image in a page is obtained. It is assumed that the text found in the vicinity of a given image on the web has high probability being relevant to the content of the image. As a result, step 112 uses keywords from the text surrounding each identified image as candidate keywords for annotating the original image. Note that some of these candidate keywords may not be relevant to the content of the original image.

[0036] The process then expands the set of keywords associated with each identified image by adding synonyms for the set of keywords (step 114). In one embodiment of the present invention, intelligent thesaurus tools are used to add synonyms for each keyword in the set. For example, keywords "sea" and "ocean" may appear in two sets of keywords for two identified images, respectively. After expanding the keywords this way, both images will be associated with both keywords.

[0037] Next, the process performs comparisons between sets of keywords for those identified images, to identify intersecting keywords (step 116). Note that adding synonyms to the keywords increases the probability of identifying such intersections.

[0038] Finally, the process takes the intersecting keywords between the multiple identified images, and annotates the original image with the intersecting keywords (step 118). Note that the relevance of the intersecting keywords to the original image can be improved by identifying more images in step 110. Because the Internet is an enormously large image database, the probability of finding better matched images increases as more and more images are searched and examined. Furthermore, the intersecting keywords obtained from more images become more accurate as well.

System [0039] FIG. 2 illustrates a computer-based system for automatically annotating an . image in accordance with an embodiment of the present invention. This computer-based system operates within a computer system 200, which can generally include any type of computer system, including, but not limited to, a computer system based on a microprocessor, a mainframe computer, a digital signal processor, a portable computing device, a personal organizer, a device controller, and a computational engine within an appliance.

[0040] During operation, computer system 200 receives an image 202 and a request to annotate image 202 with relevant textual information.

[0041] Computer system 200 includes a processor 201 and a memory 204. Memory 204 can include any type of memory that can store software modules that implement functions described in FIG. 1 for annotating image 202. As is illustrated in FIG. 2, these software modules can include: an image-receiving mechanism 205 which receives an image, an image-feature extraction mechanism 206 which extracts image features from an image; a similar-image-identification mechanism 208 which identifies images containing similar image features to input image 202; a text-obtaining mechanism which obtains text to be used to annotate an image; a matching mechanism 210 which matches an image with similar images; an annotation mechanism 211 which annotates an image; and a thesaurus and web expansion mechanism 212 which expands a list of keywords to include synonyms of the keywords. These mechanisms operate collectively to produce a set of intersecting keywords 212 which is then used to annotate image 202. [0042] Note that computer system 200 interfaces with Internet 214 to access images on Internet 214.

Application to Videos

[0043] Note that the present invention can also be used to annotate videos which do not have titles or descriptions, thereby facilitating search and retrieval operation for these videos. Specifically, a video can be partitioned into a set of representative frames, and each frame can be processed as an image. Using the above-described process, each of these representative frames can be annotated with relevant keywords. One can choose to use other existing techniques to analyze the annotations for these frames in order to generate a set of common annotations for the entire video.

[0044] The foregoing descriptions of embodiments of the present invention have been presented only for purposes of illustration and description. They are not intended to be exhaustive or to limit the present invention to the forms disclosed. Accordingly, many modifications and variations will be apparent to practitioners skilled in the art. Additionally, the above disclosure is not intended to limit the present invention. The scope of the present invention is defined by the appended claims.

Claims

What Is Claimed Is:

1. A method for automatically annotating an image, comprising: receiving the image; extracting image features from the image; identifying other images which have similar image features; obtaining text associated with the other images; identifying intersecting keywords in the obtained text associated with the other images; and annotating the image with the intersecting keywords.

2. The method of claim 1, wherein extracting the image features from the image involves: partitioning the image into tiles; and extracting the image features from the tiles.

3. The method of claim 2, wherein for each given tile size in a set of differing tile sizes, partitioning the image into tiles involves partitioning the image into tiles of the given tile size.

4. The method of claim 1, wherein prior to identifying the other images, the method further comprises matching the extracted image features against known image features in an image feature library.

5. The method of claim 4, wherein the method iiirther comprises combining the matched image features to form one or more image-feature-combinations for the image.

6. The method of claim 5, wherein identifying the other images which have similar image features involves identifying similar image-feature-combinations in the other images.

7. The method of claim 1, wherein extracting the image features from the image can involve: generating color histograms; generating orientation histograms; using a direct cosine transform (DCT) technique; using a principal component analysis (PCA) technique; or using a Gabor wavelet technique.

8. The method of claim 1, wherein the image features can be defined in terms of: shapes; colors; and textures.

9. The method of claim 1, wherein identifying the other images involves searching through images on the Internet.

10. The method of claim 1, wherein identifying the other images having the similar image features involves using probability models.

11. The method of claim 1, wherein prior to identifying the intersecting keywords, the method further comprises expanding keywords in the obtained text by adding synonyms for the keywords.

12. The method of claim 1, wherein receiving the image involves receiving the image from a video.

13. A computer-readable storage medium storing instructions that when executed by a computer cause the computer to perform a method for automatically annotating an image, the method comprising: receiving the image; extracting image features from the image; identifying other images which have similar image features; obtaining text associated with the other images; identifying intersecting keywords in the obtained text associated with the other images; and annotating the image with the intersecting keywords.

14. The computer-readable storage medium of claim 13, wherein extracting the image features from the image involves: partitioning the image into tiles; and extracting the image features from the tiles.

15. The computer-readable storage medium of claim 14, wherein for each given tile size in a set of differing tile sizes, partitioning the image into tiles involves partitioning the image into tiles of the given tile size.

16. The computer-readable storage medium of claim 13, wherein prior to identifying the other images, the method further comprises matching the extracted image features against known image features in an image feature library.

17. The computer-readable storage medium of claim 16, wherein the method further comprises combining the matched image features to form one or more image-feature-combinations for the image.

18. The computer-readable storage medium of claim 17, wherein identifying the other images which have similar image features involves identifying similar image-feature-combinations in the other images.

19. The computer-readable storage medium of claim 13, wherein extracting the image features from the image can involve: generating color histograms; generating orientation histograms; using a direct cosine transform (DCT) technique; using a principal component analysis (PCA) technique; or using a Gabor wavelet technique.

20. The computer-readable storage medium of claim 13, wherein the image features can be defined in terms of: shapes; colors; and textures.

21. The computer-readable storage medium of claim 13, wherein identifying the other images involves searching through images on the Internet.

22. The computer-readable storage medium of claim 13, wherein identifying the other images having the similar image features involves using probability models.

23. The computer-readable storage medium of claim 13, wherein prior to identifying the intersecting keywords, the method further comprises expanding keywords in the obtained text by adding synonyms for the keywords.

24. The computer-readable storage medium of claim 13, wherein receiving the image involves receiving the image from a video.

25. A computer system that automatically annotates an image, comprising: a processor; a memory; a receiving mechanism configured to receive the image; an extraction mechanism configured to extract image features from the image; an identification mechanism configured to identify other images which have similar image features; an obtaining mechanism configured to obtain text associated with the other images; wherein the identification mechanism is further configured to identify intersecting keywords in the obtained text associated with the other images; and an annotation mechanism configured to annotate the image with the intersecting keywords.

26. The computer system of claim 25, wherein the extraction mechanism is configured to: partition the image into tiles; and to extract the image features from the tiles.

27. The computer system of claim 25, further comprising a matching mechanism configured to match the extracted image features against known image features in an image feature library.

28. The computer system of claim 27, wherein the matching mechanism is configured to combine the matched image features to form one or more image-feature-combinations for the image.

29. The computer system of claim 28, wherein the identification mechanism is configured to identify similar image-feature-combinations in the other images.

30. The computer system of claim 25, wherein the identification mechanism is further configured to: search through images on the Internet; and to expand keywords in the obtained text by adding synonyms for the keywords.

31. The computer system of claim 25, wherein the receiving mechanism is configured to receive the image from a video.

32. A method for automatically annotating visual media content, comprising: receiving the visual media content; extracting features from the visual media content; identifying other visual media content which has similar features; obtaining text associated with the other visual media content; identifying intersecting keywords in the obtained text associated with the other visual media content; and annotating the visual media content with the intersecting keywords.

33. The method of claim 32, wherein extracting the features from the visual media content involves: partitioning an image within the visual media content into tiles; and extracting image features from the tiles.

34. The method of claim 33, wherein for each given tile size in a set of differing tile sizes, partitioning the image into tiles involves partitioning the image into tiles of the given tile size.

35. The method of claim 32, wherein prior to identifying the other visual media content, the method further comprises matching the extracted features against known features in a feature library.

36. The method of claim 35, wherein the method further comprises combining the matched features to form one or more feature-combinations for the visual media content.

37. The method of claim 36, wherein identifying the other visual media content which has similar features involves identifying similar feature-combinations in the other visual media content.

38. The method of claim 32, wherein extracting the features from the visual media content can involve: generating color histograms; generating orientation histograms; using a direct cosine transform (DCT) technique; using a principal component analysis (PCA) technique; or using a Gabor wavelet technique.

39. The method of claim 32, wherein the features can be defined in terms of: shapes; colors; and textures.

40. A computer-readable storage medium storing instructions that when executed by a computer cause the computer to perform a method for automatically annotating visual media content, the method comprising: receiving the visual media content; extracting features from the visual media content; identifying other visual media content which has similar features; obtaining text associated with the other visual media content; identifying intersecting keywords in the obtained text associated with the other visual media content; and annotating the visual media content with the intersecting keywords.

41. A method for automatically annotating a video, comprising:

16

XXX G:\Google\GGL-0S84-00-WO\GGL-884-00-WO Application.doc receiving the video; extracting video features from the video; identifying other videos which have similar video features; obtaining text associated with the other videos; identifying intersecting keywords in the obtained text associated with the other videos; and annotating the video with the intersecting keywords.

42. The method of claim 41, wherein extracting the video features from the video involves: partitioning an image within the video into tiles; and extracting the image features from the tiles.

43. The method of claim 42, wherein for each given tile size in a set of differing tile sizes, partitioning the image into tiles involves partitioning the image into tiles of the given tile size.

44. The method of claim 41, wherein prior to identifying the other videos, the method further comprises matching the extracted video features against known video features in a video feature library.

45. The method of claim 44, wherein the method further comprises combining the matched video features to form one or more video-feature-combinations for the video.

46. The method of claim 45, wherein identifying the other videos which have similar video features involves identifying similar video-feature-combinations in the other videos.

17

XXX G:\Google\GGW)884-00-WO\GGI^884-00-WO Application.doc

47. The method of claim 41, wherein extracting the video features from the video can involve: generating color histograms; generating orientation histograms; using a direct cosine transform (DCT) technique; using a principal component analysis (PCA) technique; or using a Gabor wavelet technique..

48. The method of claim 41, wherein the video features can be defined in terms of: shapes; colors; and textures.

49. A computer-readable storage medium storing instructions that when executed by a computer cause the computer to perform a method for automatically annotating a video, the method comprising: receiving the video; extracting video features from the video; identifying other videos which have similar video features; obtaining text associated with the other videos; identifying intersecting keywords in the obtained text associated with the other videos; and annotating the video with the intersecting keywords.

18

XXX G:\Gcogle\CKjL-0884O0-WO\GGMi84-00-\VO Application.doc