Determining dominant gradient orientation in image processing using double-angle gradients
Methods and image processing systems are provided for determining a dominant gradient orientation for a target region within an image. A plurality of gradient samples are determined for the target region, wherein each of the gradient samples represents a variation in pixel values within the target region. The gradient samples are converted into double-angle gradient vectors, and the double-angle gradient vectors are combined so as to determine a dominant gradient orientation for the target region.
1. A method of steering an anisotropic filter configured to filter pixel values in an image, the method comprising:
determining a dominant gradient orientation for a target region within the image by:
converting gradient samples for the target region into double-angle gradient vectors,
combining the double-angle gradient vectors to determine a compound gradient vector for the target region, and
converting the compound gradient vector to a dominant gradient vector, the dominant gradient vector representing the dominant gradient orientation for the target region; and
steering the anisotropic filter in dependence on the determined dominant gradient orientation.
2. The method of claim 1 , wherein the anisotropic filter uses an asymmetric filter kernel which has a minor axis and a major axis.
3. The method of claim 2 , wherein the filter kernel is an elliptical filter kernel.
4. The method of claim 2 , wherein the minor axis is the axis of the filter kernel in which samples are collected over the smallest distance, and the major axis is the axis of the filter kernel in which samples are collected over the largest distance.
5. The method of claim 2 , wherein steering the anisotropic filter comprises aligning the minor axis of the filter kernel with the determined dominant gradient orientation.
6. The method of claim 2 , wherein steering the anisotropic filter comprises aligning the minor axis of the filter kernel with the determined dominant gradient orientation so as to align the major axis of the filter kernel with the longitudinal axis of a thin-line structure in the image.
7. The method of claim 6 , wherein aligning the minor axis of the filter kernel with the determined dominant gradient orientation reduces blurring of the thin-line structure in the image during filtering.
8. The method of claim 2 , the method further comprising determining the eccentricity of the filter kernel in dependence on the magnitude of the dominant gradient vector.
9. The method of claim 8 , wherein the eccentricity of the filter kernel is larger for dominant gradient vectors with larger magnitudes, and the eccentricity of the filter kernel is smaller for dominant gradient vectors with smaller magnitudes.
10. The method of claim 9 , wherein dominant gradient vectors with larger magnitudes represent thinner line structures in the image, and dominant gradient vectors with smaller magnitudes represent thicker line structures in the image or areas of the image where there are no line structures present.
11. The method of claim 1 , the method further comprising using the steered anisotropic filter to perform edge-preserving noise reduction filtering and/or to reconstruct full colour images from mosaic images by performing de-mosaicing filtering.
12. The method of claim 1 , wherein the target region is a region surrounding a target pixel in the image.
13. The method of claim 1 , wherein each of the gradient samples for the target region represents a variation in pixel values within the target region.
14. The method of claim 1 , the method further comprising determining the gradient samples for the target region.
15. The method of claim 14 , wherein the target region is a region surrounding a target pixel, and each of the gradient samples is determined by determining a difference between: (i) the pixel value at the target pixel, and (ii) a pixel value of a neighbouring pixel positioned in a respective direction with respect to the target pixel.
16. The method of claim 1 , wherein converting the gradient samples into double-angle gradient vectors comprises encoding each gradient sample with a double-angle gradient vector that has an angle twice that of the gradient sample.
17. The method of claim 1 , wherein the dominant gradient vector has an angle half that of the compound gradient vector.
18. The method of claim 1 , wherein converting the gradient samples into double-angle gradient vectors comprises multiplying their angular components by two, and converting the compound gradient vector to the dominant gradient vector comprises dividing its angular component by two.
19. An image processing system configured to perform anisotropic filtering of pixel values in an image, the image processing system comprising:
a conversion unit configured to convert gradient samples for the target region into double-angle gradient vectors;
a combining unit configured to combine the double-angle gradient vectors to determine a compound gradient vector for the target region;
a determining unit configured to convert the compound gradient vector to a dominant gradient vector, the dominant gradient vector representing the dominant gradient orientation for the target region; and
an image processing unit configured to steer an anisotropic filter in dependence on the determined dominant gradient orientation.
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 of steering an anisotropic filter configured to filter pixel values in an image, comprising:
determining a dominant gradient orientation for a target region within the image by:
converting gradient samples for the target region into double-angle gradient vectors,
combining the double-angle gradient vectors to determine a compound gradient vector for the target region, and
converting the compound gradient vector to a dominant gradient vector, the dominant gradient vector representing the dominant gradient orientation for the target region; and
steering the anisotropic filter in dependence on the determined dominant gradient orientation.