Coverage-based image relevance ranking
View Patent ↗Implementations of coverage-based image relevance ranking are described. In one implementation, an acquired image is ranked relative to a set of previously stored images based upon the conditional entropy of the acquired image. The conditional entropy may be computed after first removing overlapping pixels that are present in both the acquired image and the set of previously stored images. Once the image is assigned a relevance rank, other decisions concerning the image may be made based on the rank, such as whether to save the image, delete the image, or use it to replace a less relevant image.
1. A method comprising:
acquiring an image having multiple pixels;
determining overlapping pixels between the acquired image and a plurality of previously acquired images;
computing a relevance ranking of the acquired image based on conditional entropy of the acquired image, with respect to an entropy contributed by the overlapping pixels and with respect to the plurality of previously acquired images; and
replacing a previously stored image with the acquired image if the relevance ranking of the acquired image is higher than that of the previously stored image.
2. A method as recited in claim 1 , wherein the acquiring an image is achieved by at least one of a camera, a satellite, a portable digital assistant, or a communication device.
3. A method as recited in claim 1 , wherein the determining overlapping pixels comprises scaling the acquired image to a predetermined size.
4. A method as recited in claim 1 , wherein the determining overlapping pixels comprises:
identifying key features from the acquired image and from the plurality of previously acquired images;
matching the key features identified from the acquired image with the key features identified from the plurality of previously acquired images; and
calculating overlap regions defined by the key features that matched in the acquired image and the plurality of previously acquired images, wherein pixels contained within the overlap regions are the overlapping pixels.
5. A method as recited in claim 4 , wherein the determining overlapping pixels further comprises eliminating the key features that are spurious.
6. A method as recited in claim 4 , wherein the identifying comprises retrieving previously stored key features of the plurality of previously acquired images.
7. A method as recited in claim 4 , further comprising:
storing the key features of the plurality of previously acquired images if the key features were not previously stored; and
storing the key features of the acquired image if the acquired image is stored.
8. A method as recited in claim 4 , wherein the computing a relevance ranking comprises:
computing a union of the overlap regions; and
removing the overlapping pixels contained within the union of overlap regions from the acquired image.
9. A method as recited in claim 1 further comprising deciding whether to store the acquired image based on the relevance ranking.
10. A computer readable medium, where the medium is not a signal, storing computer-executable instructions that, when executed, configure one or more processors to perform acts comprising:
receiving an image; and
computing a relevance ranking for the image as a function of conditional entropy of the image with respect to an entropy contributed by a plurality of previously acquired images, wherein the computing a relevance ranking comprises:
scaling the image to a predetermined size;
identifying key features of the image and the plurality of previously acquired images;
matching the key features found in both the image and the plurality of previously acquired images; and
eliminating any matching key features of the image that are spurious.
11. A computer-readable medium as recited in claim 10 , wherein the computing a relevance ranking comprises:
identifying overlapping pixels found in both the image and the plurality of previously acquired images; and
computing a relevance ranking for the image by:
computing the conditional entropy of the image, with respect to the entropy contributed by the overlapping pixels and with respect to the plurality of previously acquired images.
12. A computer-readable medium as recited in claim 10 , further comprising computer-executable instructions that, when executed by the processor, perform an additional act of deciding whether to store the image based on the relevance ranking of the image.
13. A communications device comprising:
the computer-readable medium as recited in claim 10 ; and
one or more processors to execute the computer-executable instructions stored on the computer-readable medium.
14. A device comprising:
one or more processors;
memory; and
an image analysis module, stored in the memory and executable on the one or more processors, to compute a relevance ranking of an image based on conditional entropy of the image with respect to an entropy contributed by a plurality of previously acquired images, wherein the image analysis module comprises:
a pixel analysis module to determine overlapping pixels common to both the image and the plurality of previously acquired images; and
an entropy computation module to compute the conditional entropy of the image, wherein the entropy computation module comprises:
a pixel removal module to remove the overlapping pixels from the image; and
a conditional entropy computation module to calculate the conditional entropy of the image, after the pixel removal module has removed the overlapping pixels, with respect to the plurality of previously acquired images.
15. A device as recited in claim 14 , wherein the pixel analysis module comprises:
an image scaling module to scale the image to a predetermined size;
a key feature extraction module to identify key features from the image and the plurality of previously acquired images;
a key feature matching module to match the key features identified in the image with corresponding key features identified in the plurality of previously acquired image; and
an overlap region determination module to determine overlap regions between the image and the plurality of previously acquired images, wherein the overlapping pixels are contained with the overlap regions.
16. A device as recited in claim 14 , embodied as one of at least a communication device, a camera, a computer, a satellite, a peer-to-peer network.