Stylizing geographic features in photographic images based on image content

• US 8,625,890 B1
• Filed: 10/17/2011
• Issued: 01/07/2014
• Est. Priority Date: 10/17/2011
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Brenner, Claus
• Primary Examiner(s): Wu, Jingge
• Application Number:
  US13/317,374
• Time in Patent Office: 813 Days
• Field of Search: 382/165, 382/173
• US Class Current: 382/165
• CPC Class Codes: G06T 11/001 Texturing; Colouring; Generation of texture or colour inpainting G06T5/77
• CPC Class Codes: G06T 2207/10024 Color image
• CPC Class Codes: G06T 7/90 Determination of colour characteristics

Abstract

Embodiments use the content of photographic images to stylize geographic features in the photographic images. In an embodiment, a computer-implemented method stylizes geographic features in a photographic image. In the method, a plurality of pixel regions in the photographic image are identified. For respective pixel regions, a plurality of attributes of image content within the pixel region of the photographic image are identified using the image content, and a geographic content type of the pixel region is determined using a classifier trained to recognize the geographic content type based on the plurality of attributes. Finally, at least one of the plurality of pixel regions of the photographic image are altered based on the respective determined geographic content type to stylize a geographic feature illustrated in the photographic image.

Claims

1. A computer-implemented method for stylizing geographic features in a photographic image, comprising:

(a) identifying a plurality of pixel regions in the photographic image;

for each pixel region in the plurality of pixel regions;

(b) identifying a plurality of attributes of image content within the pixel region of the photographic image using the image content, the identifying (b) comprising determining altitude values for respective pixels in the pixel region of the photographic image;

(c) normalizing the altitude values for respective pixels in the pixel region against an altitude value representing an altitude of the ground in a geographic region illustrated in the photographic image;

(d) determining a geographic content type of the pixel region using a classifier trained to recognize the geographic content type based on the plurality of attributes, the determining (d) comprising determining whether the geographic content type is a man-made structure based, at least in part, on the normalized altitude values for respective pixels, wherein a likelihood of the geographic content type being determined to be a man-made structure increases as the normalized altitude value increases; and

(e) altering at least one of the plurality of pixel regions of the photographic image based on the respective determined geographic content type to stylize a geographic feature illustrated in the photographic image.

2. The method of claim 1, wherein the identifying (b) further comprises determining a color magnitude of each of a plurality of colors illustrated within the pixel region of the photographic image, and

wherein the determining (d) comprises determining the geographic content type based, at least in part, on each color magnitude identified in (b).

3. The method of claim 2, wherein the identifying (b) further comprises determining an IR magnitude of infrared signals captured for the pixel region of the photographic image, and

wherein the determining (d) further comprises determining whether the geographic content type is vegetation based, at least in part on, on the IR magnitude and a color magnitude of a red color within the pixel region identified in (b), wherein a likelihood of the geographic content type being determined to be vegetation increases as the IR magnitude increases, and the likelihood decreases as the red magnitude increases.

4. The method of claim 1, further comprising:

(f) determining a content type mask that maps each pixel in the photographic image to a corresponding geographic content type determined in (d), and

wherein altering (e) comprises altering the at least one pixel region based on the content type mask determined in (f).

5. The method of claim 1, wherein altering (e) comprises inserting highlighting into the photographic image to emphasize a particular pixel region.

6. The method of claim 1, wherein altering (e) comprises altering the photographic image such that all pixel regions having a particular geographic content type are brightened relative to all pixel regions not having the particular geographic content type.

7. The method of claim 1, wherein altering (e) comprises altering the photographic image such that all pixel regions not having a particular geographic content type are blurred relative to all pixel regions having the particular geographic content type.

8. The method of claim 1, wherein altering (e) comprises:

(i) determining an average color of all pixel regions having a particular geographic content type; and

(ii) setting every pixel in a pixel region having the particular geographic content type to the average color determined in (i).

9. The method of claim 1, wherein determining (d) comprises selecting the geographic content type from the group consisting of roads, buildings, grass, and trees.

10. A computer-implemented system for stylizing geographic features in a photographic image, comprising:

an image region module that identifies a plurality of pixel regions in the photographic image;

a region attribute module that, for each pixel region in the plurality of pixel regions, identifies a plurality of attributes of image content within the pixel region of the photographic image based on content within the pixel region, wherein the region attribute module comprises an altitude determination module that, for the respective pixel regions;

(i) determines an altitude value for respective pixels in the pixel region of the photographic image, and

(ii) normalizes the altitude values for respective pixels in the pixel region against an altitude value representing an altitude of the ground in the geographic region illustrated in the photographic image;

a classifier module trained to recognize the geographic content type based on the plurality of attributes, wherein the classifier module determines a geographic content type for each of the respective pixel regions using a classifier, and wherein the classifier module determines whether the geographic content type is a man-made structure based, at least in part, on the normalized altitude values for respective pixels determined by the altitude determination module, wherein a likelihood of the geographic content type being determined to be a man-made structure increases as the normalized altitude value increases; and

an image alteration module that alters at least one of the plurality of pixel regions of the photographic image based on the respective determined geographic content type to stylize a geographic feature illustrated in the photographic image.

11. The system of claim 10, wherein the region attribute module determines a color magnitude of each of a plurality of colors illustrated within the pixel region of the photographic image, and

wherein the classifier module determines the geographic content type based, at east in part, on each color magnitude determined by the region attribute module.

12. The system of claim 11, wherein the region attribute module determines an IR magnitude of infrared signals captured for the pixel region of the photographic image, and

wherein the classifier module determines whether the geographic content type is vegetation based, at least in, part on, on the IR magnitude and a color magnitude of a red color within the respective pixel region, wherein a likelihood of the geographic content type being determined to be vegetation increases as the IR magnitude increases, and the likelihood decreases as the red magnitude increases.

13. The system of claim 10, further comprising:

an image mask module that determines a content type mask that maps each pixel in the photographic image to a corresponding geographic content type determined by the classifier module, and

wherein the image alteration module alters the at least one pixel region based on the content type mask.

14. The system of claim 10, wherein image alteration module inserts highlighting into the photographic image to emphasize a particular pixel region.

15. The system of claim 10, wherein image alteration module alters the photographic image such that all pixel regions having a particular geographic content type are brightened relative to all pixel regions not having the particular geographic content type.

16. The system of claim 10, wherein the image alteration, module alters the photographic image such that all pixel regions not having a particular geographic content type are blurred relative to all pixel regions having the particular geographic content type.

17. The system of claim 10, wherein image alteration module:

(i) determines an average color of all pixel regions having a particular geographic content type, and

(ii) sets every pixel in a region having the particular geographic content type to the average color.

18. The system of claim 10, wherein classifier module selects the geographic content type from the group consisting of roads, buildings, grass, and trees.

19. A computer-implemented system for stylizing geographic features in a photographic image, comprising:

a server configured to;

(i) identify a plurality of pixel regions in the photographic image, for each pixel region in the plurality of pixel regions;

(ii) identify a plurality of attributes of image content within the pixel region of the photographic image based on image content within the pixel region, the plurality of attributes including altitude values for respective pixels in the pixel region of the photographic image,

(iii) normalize the altitude values for respective pixels in the pixel region against an altitude value representing an altitude of the ground in a geographic region illustrated in the photographic image;

(iv) determine whether a geographic content type of the pixel region is a man-made structure using a classifier trained to recognize the geographic content type based on the plurality of attributes including the normalized altitude values for respective pixels such that a likelihood of the geographic content type being determined to be a man-made structure increases as the normalized altitude value increases,

(v) alter at least one of the plurality of pixel regions of the photographic image based on the respective determined geographic content type to stylize a geographic feature illustrated in the photographic image, and

(vi) provide the altered image over a network in response to a request;

a client configured to receive the altered image from the server via the network and display the image as a map.

20. A nontransitory program storage device embodying a program of instructions executable by at least one machine to perform a method for stylizing geographic features in a photographic image, said method comprising:

(a) identifying a plurality of pixel regions in the photographic image, for each pixel region in the plurality of pixel regions;

(b) identifying a plurality of attributes of image content within the pixel region of the photographic image based on image content within the pixel region, the plurality of attributes including altitude values for respective pixels in the pixel region of the photographic image,

(c) normalizing the altitude values for respective pixels in the pixel region against an altitude value representing an altitude of the ground in a geographic region illustrated in the photographic image;

(d) determining whether a geographic content type of the pixel region is a man-made structure using a classifier trained to recognize the geographic content type based on the plurality of attributes including the normalized altitude values for respective, pixels such that a likelihood of the geographic content type being determined to be a man-made structure increases as the normalized altitude value increases; and

(e) altering at least one of the plurality of pixel regions of the photographic image based on the respective determined geographic content type to stylize a geographic feature illustrated in the photographic image.