Coregistration of ground penetrating radar images
View Patent ↗A method of coregistering images is disclosed. Two dimensional images are first preprocessed, including clipping values in each image to a specified standard value range, and then applying a local equalization technique to the clipped images. Feature extraction is then performed to identify one or more features in the images. One or more of the extracted features are then matched between images to determine a degree of overlap between the images and a relationship between image coordinates of the images. The images are then transformed to a common coordinate system based on the degree of overlap and the relationship between their image coordinates. The images are then combined to form a single image.
1 . A method comprising:
accessing a first plurality of images; and
applying a coregistration process to a second plurality of images in the first plurality of images, the coregistration process including:
converting the second plurality of images to a specified standard form, by:
for each image of the second plurality of images, computing a mean and a variance of pixel values in the image; and
for each pixel value in each image of the second plurality of images, subtracting the mean computed for that image from each pixel value in the image to produce an intermediate pixel value, and dividing each intermediate pixel value by a square root of the variance computed for that image to produce a standardized pixel value;
preprocessing the second plurality of images, wherein the preprocessing includes clipping values in each image of the second plurality of images to a
specified standard value range, to produce a plurality of clipped images; and applying a local equalization technique to the plurality of clipped images;
performing a feature extraction to identify a feature in each image of the plurality of clipped images;
matching the feature between each image of the plurality of clipped images, to determine a degree of overlap between the images in the second plurality of images and a relationship between image coordinates of the images in the second plurality of images;
transforming the second plurality of images to a common coordinate system based on the degree of overlap and the relationship between the image coordinates of the second plurality of images; and
combining the second plurality of images to form a single image.
2 . The method of claim 1 , wherein each of the first plurality of images is a two-dimensional subset or projection of a three-dimensional image.
3 . The method of claim 1 , wherein the local equalization technique comprises a local histogram equalization in which image values are adjusted within local regions to have a flat histogram.
4 . The method of claim 1 , wherein the preprocessing further includes, prior to the clipping values, converting the second plurality of images to a specified standard form, by converting the second plurality of images to a common brightness scale.
5 . The method of claim 1 , wherein the combining the second plurality of images comprises summing overlapping portions of the images.
6 . The method of claim 1 , wherein the feature comprises a feature descriptor vector and a location.
7 . The method of claim 1 , wherein the matching a common feature present in each image of the plurality of clipped images comprises minimizing a distance between feature descriptor vectors between images.
8 . The method of claim 1 , wherein the feature extraction comprises a histogram of oriented gradients (HOG) technique.
9 . The method of claim 1 , wherein the feature extraction comprises using at least one of:
an edge oriented histogram;
scale invariant feature transform descriptors; or
edge histogram descriptors.
10 . The method of claim 1 , wherein the first plurality of images are ground penetrating radar (GPR) images.
11 . A computer system comprising:
one or more processor; and
one or more memories coupled to the one or more processors and storing instructions, execution of which by the one or more processors causes the computer system to:
access a first plurality of images acquired by a ground penetrating radar (GPR) system; and
apply a coregistration process to a second plurality of images in the first plurality of images, the coregistration process including:
preprocessing the second plurality of images, wherein the preprocessing includes
converting the second plurality of images to a specified standard form, by:
for each image of the second plurality of images, computing a mean and a variance of pixel values in the image; and
for each pixel value in each image of the second plurality of images, subtracting the mean computed for that image from each pixel value in the image to produce an intermediate pixel value, and dividing each intermediate pixel value by a square root of the variance computed for that image to produce a standardized pixel value;
clipping values in each image of the second plurality of images to a specified standard value range, to produce a plurality of clipped images; and
applying a local equalization technique to the plurality of clipped images;
performing a feature extraction to identify a feature in each image of the plurality of clipped images;
matching the feature between each image of the plurality of clipped images, to determine a degree of overlap between the images in the second plurality of images and a relationship between image coordinates of the images in the second plurality of images;
transforming the second plurality of images to a common coordinate system based on the degree of overlap and the relationship between the image coordinates of the second plurality of images; and
combining the second plurality of images to form a single image.
12 . The computer system of claim 11 , wherein each of the first plurality of images is a two-dimensional subset or projection of a three-dimensional image.
13 . The computer system of claim 11 , wherein the local equalization technique comprises a local histogram equalization in which image values are adjusted within local regions to have a flat histogram.
14 . The computer system of claim 11 , wherein the preprocessing further includes, prior to the clipping values, converting the second plurality of images to a specified standard form, by converting the second plurality of images to a common brightness scale.
15 . The computer system of claim 11 , wherein the combining the second plurality of images comprises summing overlapping portions of the images.
16 . The computer system of claim 11 , wherein the feature comprises a feature descriptor vector and a location.
17 . The computer system of claim 11 , wherein the matching a common feature present in each image of the plurality of clipped images comprises minimizing a distance between feature descriptor vectors between images.
18 . The computer system of claim 11 , wherein the feature extraction comprises a histogram of oriented gradients (HOG) technique.
19 . The computer system of claim 11 , wherein the feature extraction comprises using at least one of:
an edge oriented histogram;
scale invariant feature transform descriptors; or
edge histogram descriptors.
20 . The computer system of claim 11 , wherein the first plurality of images have been acquired by use of a linear GPR array.
21 . A non-transitory machine-readable storage medium storing instructions, execution of which by one or more processors in a processing system causes the processing system to perform a process comprising:
accessing a first plurality of images acquired by a ground penetrating radar (GPR) system; and
applying a coregistration process to a second plurality of images in the first plurality of images, the coregistration process including:
converting the second plurality of images to a specified standard form, by:
for each image of the second plurality of images, computing a mean and a variance of pixel values in the image; and
for each pixel value in each image of the second plurality of images, subtracting the mean computed for that image from each pixel value in the image to produce an intermediate pixel value, and dividing each intermediate pixel value by a square root of the variance computed for that image to produce a standardized pixel value;
preprocessing the second plurality of images, wherein the preprocessing includes clipping values in each image of the second plurality of images to a
specified standard value range, to produce a plurality of clipped images; and applying a local equalization technique to the plurality of clipped images;
performing a feature extraction to identify a feature in each image of the plurality of clipped images;
matching the feature between each image of the plurality of clipped images, to determine a degree of overlap between the images in the second plurality of images and a relationship between image coordinates of the images in the second plurality of images;
transforming the second plurality of images to a common coordinate system based on the degree of overlap and the relationship between the image coordinates of the second plurality of images; and
combining the second plurality of images to form a single image.
22 . The non-transitory machine-readable storage medium of claim 21 , wherein each of the first plurality of images is a two-dimensional subset or projection of a three-dimensional image.
23 . The non-transitory machine-readable storage medium of claim 21 , wherein the local equalization technique comprises a local histogram equalization in which image values are adjusted within local regions to have a flat histogram.
24 . The non-transitory machine-readable storage medium of claim 21 , wherein the preprocessing further includes, prior to the clipping values, converting the second plurality of images to a specified standard form, by converting the second plurality of images to a common brightness scale.
25 . The non-transitory machine-readable storage medium of claim 21 , wherein the combining the second plurality of images comprises summing overlapping portions of the images.
26 . The non-transitory machine-readable storage medium of claim 21 , wherein the feature comprises a feature descriptor vector and a location.
27 . The non-transitory machine-readable storage medium of claim 21 , wherein the matching a common feature present in each image of the plurality of clipped images comprises minimizing a distance between feature descriptor vectors between images.
28 . The non-transitory machine-readable storage medium of claim 21 , wherein the feature extraction comprises a histogram of oriented gradients (HOG) technique.
29 . The non-transitory machine-readable storage medium of claim 21 , wherein the feature extraction comprises using at least one of:
an edge oriented histogram;
scale invariant feature transform descriptors; or
edge histogram descriptors.
30 . The non-transitory machine-readable storage medium of claim 21 , wherein the first plurality of images have been acquired by use of a linear GPR array.