Identifying features
A method for identifying features in an image, the method comprising: a) from a first location, scanning the image along a scan path in a scan direction away from the first location so as to detect a border in the image at a second location; b) mapping a path defined by the border in a first direction along the border from the second location, the path being mapped in dependence on a border quality measure; c) determining whether the mapped path forms a perimeter; and d) if the mapped path does form a perimeter, scanning the image from the mapped path so as to detect a further border in the image; or if the mapped path does not form a perimeter, mapping a path defined by the border from the second location in a second direction along the border that is different to the first direction.
1. A method for identifying features in an image, the method comprising:
a) from a first location, scanning the image along a single scan path in a scan direction away from the first location so as to detect a border in the image at a second location;
b) mapping a path defined by the border in a first direction along the border from the second location, the path being mapped in dependence on a border quality measure;
c) determining whether the mapped path forms a perimeter; and
d) if the mapped path does form a perimeter, scanning the image from the mapped path so as to detect a further border in the image; and if the mapped path does not form a perimeter, mapping a path defined by the border from the second location in a second direction along the border that is different to the first direction.
2. A method as claimed in claim 1 comprising:
on mapping said path in the second direction, scanning the image from the mapped path so as to detect a further border in the image.
3. A method as claimed in claim 1 , wherein the method comprises, on detecting a further border in the image at a third location, repeating steps b) to d), wherein the repeated steps are performed from the third location.
4. A method as claimed in claim 1 , wherein the method comprises, after having traced a border:
determining whether the circumference or length of the border and/or the compounded quality measurements of the border exceeds a predetermined threshold; and if not, discarding the border.
5. A method as claimed in claim 1 , wherein the method comprises, from a fourth location: scanning the image along a second scan path in the scan direction away from the fourth location so as to detect a border in the image at a fifth location, the fourth location being offset by a predetermined distance from the first location in a direction perpendicular to the scan direction.
6. A method as claimed in claim 1 , the method comprising, on detecting the border:
ignoring the border if it has been mapped previously.
7. A method as claimed in claim 1 , wherein in each step of scanning the image, the scan direction is the same.
8. A method as claimed in claim 1 , wherein the image is composed of a plurality of pixels and the method comprises mapping a path by detecting borders between pixel positions, wherein the quality measure is indicative of the quality of the border at each border position.
9. A method as claimed in claim 8 , wherein the method comprises:
estimating an attribute for each of a set of two adjacent pixel positions in the image;
assessing whether a predetermined binary condition differs in respect of the two pixel positions, and if so, determining that a border is present in the part of the image represented by the two pixel positions.
10. A method as claimed in claim 9 , wherein the border quality measure is a value derived from neighbouring pixels, the method further comprises: determining whether the derived value is above or below a predetermined threshold, and if above, mapping that border position as part of the path; or if below, not mapping that border position as part of the path.
11. A method as claimed in claim 10 , wherein the attribute of the pixel position is based on a Laplacian of Gaussian filter applied to the image and centred on that pixel position.
12. A method as claimed in claim 9 , wherein the attribute of a pixel position is based on estimating the second-order derivative of an isotropic Gaussian convolution of the image at the pixel position and with respect to the gradient direction of the Gaussian.
13. A method as claimed in claim 12 , wherein a second attribute of a pixel position is based on estimating the third-order derivative of an isotropic Gaussian convolution of the image at the pixel position and with respect to the gradient direction of the Gaussian, wherein the border quality measure at a given border position is an interpolation of the second attribute with respect to each of the two pixel positions on either side of the border position.
14. A method as claimed in claim 1 , wherein the image is a frame of a video stream.
15. A device comprising:
a memory storing non-transient program code; and
a processor configured to identify features in an image by executing the code to perform the steps of:
a) from a first location, scanning the image along a scan path in a single scan direction away from the first location so as to detect a border in the image at a second location;
b) mapping a path defined by the border in a first direction along the border from the second location, the path being mapped in dependence on a border quality measure;
c) determining whether the mapped path forms a perimeter; and
d) if the mapped path does form a perimeter, scanning the image from the mapped path so as to detect a further border in the image; and if the mapped path does not form a perimeter, mapping a path defined by the border from the second location in a second direction along the border that is different to the first direction.
16. A device as claimed in claim 15 , wherein the device comprises a camera and the image is an image captured by the camera.
17. A device as claimed in claim 16 , wherein the image is a frame of a video stream captured by the camera.
18. A device as claimed in claim 17 , wherein the device is configured to perform the said steps in real time on successive frames of the video stream.
19. A device as claimed in claim 15 , wherein the processor is further configured to:
estimate an attribute for each of a set of two adjacent pixel positions in the image;
assess whether a predetermined binary condition differs in respect of the two pixel positions, and if so, determine that a border is present in the part of the image represented by the two pixel positions.
20. A non-transitory computer readable storage medium having stored thereon computer readable instructions that, when executed at a computer system, cause the computer system to perform a method for identifying features in an image, the method comprising:
a) from a first location, scanning the image along a scan path in a single scan direction away from the first location so as to detect a border in the image at a second location;
b) mapping a path defined by the border in a first direction along the border from the second location, the path being mapped in dependence on a border quality measure;
c) determining whether the mapped path forms a perimeter; and
d) if the mapped path does form a perimeter, scanning the image from the mapped path so as to detect a further border in the image; and if the mapped path does not form a perimeter, mapping a path defined by the border from the second location in a second direction along the border that is different to the first direction.