Systems and methods for generating depth maps using a set of images containing a baseline image
View Patent ↗Systems and methods for implementing array cameras configured to perform super-resolution processing to generate higher resolution super-resolved images using a plurality of captured images and lens stack arrays that can be utilized in array cameras are disclosed. An imaging device in accordance with one embodiment of the invention includes at least one imager array, and each imager in the array comprises a plurality of light sensing elements and a lens stack including at least one lens surface, where the lens stack is configured to form an image on the light sensing elements, control circuitry configured to capture images formed on the light sensing elements of each of the imagers, and a super-resolution processing module configured to generate at least one higher resolution super-resolved image using a plurality of the captured images.
1. A method of generating a depth map from a plurality of images captured from different viewpoints, comprising:
normalizing a set of images using calibration data stored in a storage device using an address conversion module, where:
the set of images comprises a plurality of images that are:
captured from different viewpoints including sampling diversity;
include different occlusions sets;
wherein the occlusion set of a first image is the portion of a scene visible in a second image that is occluded in the first image; and
the set of images includes:
images captured using the same imaging characteristics;
images captured using different imaging characteristics; and
a first image that captures pixels around an edge of a foreground object that is visible in a baseline image and is in the occlusion set of a second image captured using the same imaging characteristics as the first image;
detecting and metering parallax using a parallax confirmation and measurement module, where detecting and metering parallax comprises:
determining the parallax that yields the highest correlation between pixels from images captured using the same imaging characteristics accounting for the positions of imagers that captured the images by:
performing redundant pair-wise measurements between pixels from images captured using the same imaging characteristics to determine pixel correlation for different parallax-induced shifts;
keeping track of various pair-wise measurements and calculating a parallax difference that yields the highest pixel correlation ignoring pixels in the images that are in an exposed occlusion set; and
generating a depth map using the parallax confirmation and measurement module based upon the measured parallax.
2. The method of claim 1 , wherein:
the calibration data indicates mappings between addresses of physical pixels in imagers and logical addresses within an image; and
the address conversion module resolves phase differences between the set of images by converting addresses of physical pixels in imagers to logical addresses.
3. The method of claim 1 , further comprising performing temperature normalization using an upstream pipeline processing module based upon a temperature measurement.
4. The method of claim 3 , wherein performing temperature normalization comprises performing a refractive index normalization using a stored point spread function selected based upon the temperature measurement.
5. The method of claim 1 , wherein calculating a parallax difference that yields the highest pixel correlation comprises determining parallax based upon the set of pair-wise measurements that have the best least squares fit.
6. The method of claim 1 , further comprising:
aligning portions of different images to compensate for parallax using an image pixel correlation module based upon the measured parallax and the stored calibration data; and
obtaining a higher resolution image having a resolution that is higher than the resolutions of the images in the set of images using a super-resolution module, where color information around the edge of the foreground object that is visible to the baseline image and in the occlusion set of the second image is reconstructed in the higher resolution image using the pixels captured by the first image.
7. The method of claim 6 , wherein obtaining a higher resolution image having a resolution that is higher than the resolutions of the images in the set of images comprises fusing the aligned image portions.
8. The method of claim 7 , wherein obtaining a higher resolution image having a resolution that is higher than the resolutions of the images in the set of images comprises performing super-resolution processing on the fused image portions to synthesize a super-resolution image.
9. The method of claim 6 , wherein obtaining a higher resolution image having a resolution that is higher than the resolutions of the images in the set of images comprises performing super-resolution processing on the aligned image portions to synthesize a super-resolved image.
10. The method of claim 6 , further comprising selecting at least one distance as a focal plane and applying blurring to pixels in the higher resolution image with depths in the depth map that are not proximate a focal plane using the super-resolution processing module.
11. The method of claim 1 , further comprising estimating distance to an object visible in the set of images using the parallax confirmation and measurement module.
12. The method of claim 1 , further comprising measuring the absolute size and shape of a physical item visible in the set of images using the parallax confirmation and measurement module.
13. The method of claim 1 , wherein the set of images is captured by an array of between 2×2 and 6×6 imagers.
14. The method of claim 1 , wherein the set of images is captured by a set of imagers that form at least a 3×3 configuration of imagers comprising:
an imager having a green filter at the center of a 3×3 configuration of imagers;
two imagers having red filters located on opposite sides of the 3×3 configuration of imagers;
two imagers having blue filters located on opposite sides of the 3×3 configuration of imagers; and
four imagers having green filters surrounding the central imager having a green filter.
15. The method of claim 1 , wherein the set of images is captured by a linear array of imagers.
16. The method of claim 1 , wherein each image in the set of images is obtained by an imager including a filter selected from the group consisting of a Bayer filter, one or more Blue filters, one or more Green filters, one or more Red filters, one or more shifted spectral filters, one or more near-IR filters, and one or more hyper-spectral filters.
17. The method of claim 1 , wherein the images in the set of images have the same resolution.
18. The method of claim 1 , wherein the images in the set of images have different resolutions.
19. A method of generating a depth map from a plurality of images captured from different viewpoints, comprising:
normalizing a set of images using calibration data stored in a storage device using an address conversion module, where:
the set of images comprises a plurality of images that are:
captured from different viewpoints including sampling diversity;
include different occlusions sets;
wherein the occlusion set of a first image is the portion of a scene visible in a second image that is occluded in the first image; and
the set of images includes:
images captured using the same imaging characteristics; and
images captured using different imaging characteristics;
detect and meter parallax using a parallax confirmation and measurement module, where detecting and metering parallax comprises:
determining the parallax that yields the highest correlation between pixels from images captured using the same imaging characteristics accounting for the positions of the imagers that captured the images by:
performing redundant pair-wise measurements between pixels from images captured using the same imaging characteristics to determine pixel correlation for different parallax-induced shifts;
keeping track of various pair-wise measurements and calculating a parallax difference that yields the highest pixel correlation ignoring pixels in the images that are in an exposed occlusion set; and
aligning portions of different images to compensate for parallax using an image pixel correlation module based upon the measured parallax and the stored calibration data;
obtaining a higher resolution image having a resolution that is higher than the resolutions of the images in the set of images using a super-resolution module, where color information around the edge of the foreground object that is visible to the baseline image and in the occlusion set of the second image is reconstructed in the higher resolution image using the pixels captured by the first image; and
generating a depth map for the higher resolution image using the parallax confirmation and measurement module based upon the measured parallax;
wherein obtaining an image having a resolution that is higher than the resolutions of the images in the set of images comprises:
fusing the aligned image portions;
performing super-resolution processing on the fused image portions to synthesize a super-resolution image; and
selecting at least one distance as a focal plane and applying blurring to pixels in the higher resolution image with depths in the depth map that are not proximate a focal plane.
20. The method of claim 6 , wherein aligning portions of different images to compensate for parallax using an image pixel correlation module further comprises determining appropriate X and Y offsets to be applied to logical pixel address calculations using an address conversion module based upon the detected and metered parallax and the stored calibration data.