Extraction of image feature data from images
An apparatus and method for obtaining image feature data of an image are disclosed herein. A color histogram of the image is extracted from the image, the extraction of the color histogram including performing one-dimensional sampling of pixels comprising the image in each of a first dimension of a color space, a second dimension of the color space, and a third dimension of the color space. An edge map corresponding to the image is analyzed to detect a pattern included in the image. In response to a confidence level of the pattern detection being below a pre-defined threshold, extracting from the image an orientation histogram of the image. And identify a dominant color of the image.
1. A method comprising:
generating a color histogram of an image by performing one-dimensional samplings of pixels of the image in each dimension of a multi-dimensional color space;
performing pattern detection on an edge map of the image;
generating an orientation histogram of the image in response to a confidence level of the pattern detection failing to transgress a threshold value, the generating of the orientation histogram being performed by a processor of a machine; and
determining a dominant color of the image.
2. The method of claim 1 , wherein:
the multi-dimensional color space includes a hue dimension, a saturation dimension, and a value dimension.
3. The method of claim 1 , wherein:
the multi-dimensional color space includes a first dimension, a second dimension, and a third dimension; and
the performing of the one-dimensional samplings includes sampling a first pixel among the pixels of the image in the first dimension in 24 bins, sampling the first pixel in the second dimension in 8 bins, and sampling the first pixel in the third dimension in 8 bins.
4. The method of claim 3 , wherein:
a second pixel among the pixels of the image has a saturation that fails to transgress a predetermined threshold; and
the performing of the one-dimensional samplings includes sampling the second pixel in a fourth dimension distinct from the first, second, and third dimensions of the multi-dimensional color space.
5. The method of claim 1 , wherein:
the multi-dimensional color space includes a first dimension, a second dimension, and a third dimension; and
the generating of the color histogram includes applying a first weight to a one-dimensional sampling of a first pixel among the pixels of the image in the first dimension, a second weight to a one-dimensional sampling of the first pixel in the second dimension, and a third weight to a one-dimensional sampling of the first pixel in the third dimension.
6. The method of claim 5 , wherein:
a second pixel among the pixels of the image has a saturation that fails to transgress a predetermined threshold; and
the generating of the color histogram includes applying a fourth weight to a one-dimensional sampling of the second pixel in a fourth dimension distinct from the first, second, and third dimensions of the multi-dimensional color space.
7. The method of claim 5 , wherein:
the generating of the color histogram includes combining a plurality of weighted one-dimensional samples of the pixels of the image to form the color histogram.
8. The method of claim 1 , wherein:
the determining of the dominant color of the image determines that the dominant color is present on more spatial area within a sample area of the image than other colors within the sample area of the image.
9. The method of claim 1 , wherein:
the determining of the dominant color of the image includes clustering colors of a sample area of the image and identifying a largest color cluster among a plurality of colors of the sample area of the image.
10. The method of claim 9 , wherein:
the multi-dimensional color space is a first multi-dimensional color space; and
the determining of the dominant color of the image includes clustering the colors of the sample area in accordance with a second multi-dimensional color space different from the first multi-dimensional color space.
11. The method of claim 10 , wherein:
the second multi-dimensional space is a nonlinear color space adopted by the International Commission on Illumination.
12. The method of claim 1 , wherein:
the performing of the one-dimensional samplings of the pixels includes performing uniform one-dimensional samplings of the pixels.
13. The method of claim 1 , wherein:
the performing of the pattern detection includes calculating a circularity value of a blob within the edge map of the image.
14. The method of claim 1 further comprising:
determining that the orientation histogram is indicative of the image having low spatial variation; and wherein
the generating of the color histogram is in response to the orientation histogram indicating that the image has low spatial variation.
15. The method of claim 1 further comprising:
determining that the orientation histogram is indicative of the image having high spatial variation;
determining a confidence score of an orientation of the image; and
indexing the image in accordance with the confidence score of the orientation of the image.
16. The method of claim 1 , wherein:
the image depicts a sellable item.
17. A non-transitory machine-readable storage medium comprising instructions that, when executed by one or more processors of a machine, cause the machine to perform operations comprising:
generating a color histogram of an image by performing one-dimensional samplings of pixels of the image in each dimension of a multi-dimensional color space;
performing pattern detection on an edge map of the image;
generating an orientation histogram of the image in response to a confidence level of the pattern detection failing to transgress a threshold value; and
determining a dominant color of the image.
18. The non-transitory machine-readable storage medium of claim 17 , wherein:
the generating of the orientation histogram includes:
calculating an x-derivative and a y-derivative of the edge map;
calculating a gradient and an orientation based on the x-derivative and the y-derivative; and
applying a weight to each edge pixel in the edge map to obtain the orientation histogram.
19. A system comprising:
one or more processors;
a color histogram module that configures at least one processor among the one or more processors to generate a color histogram of an image by performing one-dimensional samplings of pixels of the image in each dimension of a multi-dimensional color space;
a pattern module that configures at least one processor among the one or more processors to perform pattern detection on an edge map of the image;
an orientation histogram module that configures at least one processor among the one or more processors to generate an orientation histogram of the image in response to a confidence level of the pattern detection failing to transgress a threshold value; and
a dominant color module that configures at least one processor among the one or more processors to determine a dominant color of the image.
20. The system of claim 19 further comprising:
a capture module configured to receive the image from a device communicatively coupled to the system.