Alignment of lens array images using autocorrelation
In a parallax correction method and system, corresponding regions of first and second digital images of the same subject are upsampled to provide a pair of upsampled regions. The corresponding regions have parallax. The upsampled regions are statistically compared at a plurality of relative subpixel offsets to provide a corresponding plurality of similarity values. The similarity values define a peak similarity value and a corresponding peak subpixel offset among the plurality of relative subpixel offsets. The second digital image is shifted relative to the first digital image by the peak subpixel offset.
1. A parallax correction method comprising the steps of:
upsampling corresponding regions of first and second digital images of the same scene to provide a pair of upsampled regions, said corresponding regions having parallax;
statistically comparing said upsampled regions at a plurality of relative subpixel offsets to provide a corresponding plurality of similarity values, said similarity values defining a peak similarity value and a corresponding peak subpixel offset among said plurality of relative subpixel offsets; and
shifting said second digital image relative to said first digital image by said peak subpixel offset.
2. The method of claim 1 wherein each of said subpixel offsets is in the same direction.
3. The method of claim 1 wherein said shifting further comprises interpolating said first and second digital images in accordance with said peak subpixel offset.
4. The method of claim 1 wherein said comparing further comprises autocorrelating said subpixel images at said plurality of relative subpixel offsets and said peak similarity value is a peak autocorrelation value.
5. The method of claim 1 further comprising repeating said upsampling, comparing, and shifting steps for said first digital image and each of a plurality of additional digital images.
6. The method of claim 1 further comprising shifting each of a plurality of additional digital images relative to said first digital image and proportional to said peak subpixel offset.
7. The method of claim 1 further comprising:
capturing a light image with each of a plurality of sections of a capture unit to provide a plurality of captured images, each of said sections being aligned with a corresponding one of an array of lenses;
converting said plurality of captured images into a corresponding plurality of digital images, which includes said first and second digital images.
8. A parallax correction method comprising the steps of:
upsampling corresponding regions of first and second digital images of the same subject to provide a pair of upsampled regions, said corresponding regions having parallax;
statistically comparing said upsampled regions at a plurality of relative subpixel offsets to provide a corresponding plurality of similarity values, said similarity values defining a peak similarity value and a corresponding peak subpixel offset among said plurality of relative subpixel offsets; and
shifting said second digital image relative to said first digital image by said peak subpixel offset, wherein
said subpixel offsets each have a magnitude that is an integer multiple of a subpixel unit dimension defined by said upsampling.
9. A parallax correction method comprising the steps of:
directing a plurality of light images of the same scene to a corresponding section of an imager via an array of individual lenses;
providing each section of the imager with separate grid of pixels for capturing the plurality of light images of the same scene, each of the plurality of light images representing a different perspective of the same scene propagated through each of the individual lenses;
upsampling corresponding regions of the plurality of light images representing a different perspective of the same scene from different sections of the imager scene to provide a plurality of upsampled regions of the same scene, said corresponding regions having parallax;
statistically comparing said upsampled regions at a plurality of relative subpixel offsets to provide a corresponding plurality of similarity values, said similarity values defining a peak similarity value and a corresponding peak subpixel offset among said plurality of relative subpixel offsets; and
shifting said second digital image relative to said first digital image by said peak subpixel offset.
10. The method of claim 9 wherein each of said subpixel offsets is in the same direction.
11. The method of claim 9 wherein said shifting further comprises the step of interpolating said first and second digital images in accordance with said peak subpixel offset.
12. The method of claim 9 wherein said comparing further comprises autocorrelating said subpixel images at said plurality of relative subpixel offsets and said peak similarity value is a peak autocorrelation value.
13. The method of claim 9 further comprising repeating said upsampling, comparing, and shifting steps for said first digital image and each of a plurality of additional digital images.
14. The method of claim 9 further comprising shifting each of a plurality of additional digital images relative to said first digital image and proportional to said peak subpixel offset.