Identifying points of interest in an image
View Patent ↗Points of interest are identified in an image to characterize that image by dividing the image tiles consisting of adjacent pixels; finding within each tile the position of a pixel with an extremum value and ascribing that extremal value to the tile; identifying a tile with an extremal value which is more extreme than that of all adjacent tiles; and selecting as a point of interest the position within the image of the pixel with the extremum value in that identified tile.
1. A method of searching images, the method comprising the steps of:
in an image processor selecting a plurality of points of interest;
in which each said point of interest is represented by one or more co-ordinates of a pixel;
in which a first set of points of interest from a first image A are tested for matching against a second set of points of interest from a second image B, by
selecting a candidate point P 0 from the image A;
selecting a candidate match L 0 from the image B;
selecting a second different candidate point P 1 from the image A;
using the geometric relationship between P 0 and P 1 to define a search space for candidate points L 1 in the image B;
for any viable L 1 candidate points, using the pairs of points, (P 0 , P 1 ) and (L 0 , L 1 ) to define an assumed model to be fitted between the two images;
selecting a third different candidate point P 2 from the image A;
using said model to predict the position of a candidate match L 2 in the image B; and
testing for the presence of a feature point at location L 2 in the image B.
2. The method according to claim 1 where the said image is a frame of video data and the pixel values are related to luminance or colour values.
3. The method according to claim 2 in which tiles at the edge of the frame are disregarded.
4. The method according to claim 1 in which the representation of each said point of interest includes a prominence parameter.
5. The method according to claim 4 in which the said prominence parameter is a measure of the amount that the value of a pixel differs from the average pixel value for the pixels of the tile that includes that pixel.
6. The method according to claim 4 in which points of interest having low prominence are discarded.
7. The method according to claim 1 in which the said image is divided into a plurality of regions and the said image is characterised by at least one interest point in each region.
8. The method according to claim 1 in which pixel values are low-pass filtered prior to the identification of the said point of interest.
9. The method according to claim 1 , wherein the step of using the geometric relationship between P 0 and P 1 to define a search space for candidate points L 1 in image B comprises using at least one defined constraint on the minimum or maximum scaling differences between the image A and image B to define a reduced search space for candidate points L 1 in image B.
10. The method according to claim 1 , wherein the step of selecting a candidate match L 0 from image B is subject to a minimum implied overlap in translation between image A and image B.
11. The method according to claim 1 , in which a search space is reduced by imposing a constraint selected from the group consisting of an aspect ratio change limit and a temporal change limit.
12. The method according to claim 1 in which the step of testing for the presence of a feature point at location L 2 in image B has a predefined tolerance.
13. A non-transitory computer readable medium comprising a computer program product storing computer instructions, which when executed by a computer implements a method of searching images comprising the steps of:
selecting a plurality of points of interest,
in which each said point of interest is represented by one or more co-ordinates of a pixel;
in which a first set of points of interest from a first image A are tested for matching against a second set of points of interest from a second image B, by
selecting a candidate point P 0 from the image A:
selecting a candidate match L 0 from the image B:
selecting a second different candidate point P 1 from the image A;
using the geometric relationship between P 0 and P 1 to define a search space for candidate points L 1 in the image B;
for any viable L 1 candidate points, using the pairs of points, (P 0 , P 1 ) and (L 0 , L 1 ) to define an assumed model to be fitted between the two images;
selecting a third different candidate point P 2 from the image A;
using said model to predict the position of a candidate match L 2 in the image B; and
testing for the presence of a feature point at location L 2 in image B.
14. The non-transitory computer readable medium according to claim 13 where the said image is a frame of video data and the pixel values are related to luminance or colour values.
15. The non-transitory computer readable medium according to claim 13 in which tiles at the edge of the frame are disregarded.
16. A method for processing a first image in accordance with image boundaries of a matched second image, the method comprising the steps of:
analysing a first set of coordinate points from a first image having a first image boundary with a second set of coordinate points from a second image having a second image boundary, the sets of coordinate points to identify a match, wherein relative positions of matched points from respective images are used to infer relative positions of the respective image boundaries; and
processing the first image in accordance with said relative positions;
in which a candidate match between sets implied by the initial matching of two points in each group, is used to predict the relative coordinate scaling between the two sets of points, in order to predict and test the correspondence of subsequent pairs of points in the respective sets; and
in which the predicted relative scaling of a two point match between sets is used to reject potential third point matches that have a match error above a given threshold, and thus reduce the search space.
17. A method for processing a first image in accordance with image boundaries of a matched second image, the method comprising the steps of:
analysing a first set of coordinate points from a first image having a first image boundary with a second set of coordinate points from a second image having a second image boundary, the sets of coordinate points to identify a match, wherein relative positions of matched points from respective images are used to infer relative positions of the respective image boundaries; and
processing the first image in accordance with said relative positions;
in which a candidate match between sets implied by the initial matching of two points in each group, is used to predict the relative coordinate scaling between the two sets of points, in order to predict and test the correspondence of subsequent pairs of points in the respective sets; and
in which the predicted relative scaling of a two point match between sets is used to reject potential fourth point matches that have a match error above a given threshold so as to and reduce the search space.