Cloud removal by illumination normalization and interpolation weighted by cloud probabilities

Patent metadata

Field	Value
Publication number	US-12555350-B2
Application number	US-202318181183-A
Country	US
Kind code	B2
Filing date	Mar 9, 2023
Priority date	Oct 6, 2022
Publication date	Feb 17, 2026
Grant date	Feb 17, 2026

Abstract

Clouds in a satellite image are replaced with a prediction of what was occluded by those clouds. The cloudy portion of the image is interpolated from a series of satellite images taken over time, some of which are cloud-free in the target image's cloudy portion. In some configurations, clouds are removed taking into account each pixel's availability—a measure of certainty that a pixel is cloud-free. Furthermore, these images may have been taken under different amounts of illumination, making it difficult to determine whether a difference between two images is due to a change in illumination or a change to the location. The effect of illumination on each image is removed before interpolating the cloudy portion of the image. In some configurations, removing the effect of illumination also takes into account pixel availability.

First claim

What is claimed is:

1 . A method comprising:
determining an illumination for each of a series of images captured by a satellite based on a cloud-weighted average of illumination values; modifying one or more color values of one of the series of images captured by the satellite to remove an effect of illumination; determining a portion of the one of the series of images is a cloud; and generating a prediction of what was occluded by the cloud in the portion of the one of the series of images based in part on the modified color values.

2 . The method of claim 1 , wherein the series of images captured by the satellite comprises a four-dimensional image, wherein the four dimensions include an X-coordinate, a Y-coordinate, a color spectrum index, and a time, wherein the four-dimensional image is stored as a two-dimensional matrix, wherein the first dimension of the two-dimensional matrix combines the color spectrum index and the time, and wherein the second dimension of the two-dimensional matrix combines the X-coordinate and the Y-coordinate.

3 . The method of claim 2 , wherein the prediction of what was occluded by the cloud is generated by minimizing differences between color values of the series of images and color values of corresponding predictions of the series of images.

4 . The method of claim 3 , wherein the differences are minimized by minimizing, for cloud-free pixels, a square root of a sum of squares of differences between actual color values of pixels of the series of images and corresponding color values of pixels of the predictions of the series of images.

5 . The method of claim 2 , wherein the prediction of what was occluded by the cloud is made by minimizing changes in color values of predictions of images that are adjacent in the series of images.

6 . The method of claim 5 , wherein the changes in color are minimized by minimizing a square root of a sum of squares of differences between color values of pixels of adjacent predictions of the series of images.

7 . The method of claim 6 , wherein a constant value multiplied by the minimized square root determines a relative importance of minimizing a rate of change in color values in the predictions of the series of images and a minimization of a difference between the color values in adjacent predictions of the series of images.

8 . A computer-readable storage medium having computer-executable instructions stored thereupon which, when executed by a computing device, cause the computing device to:
determine an illumination level for each of a series of images captured by a satellite based on a cloud-weighted average of illumination values; modify color values of one of the series of images captured by the satellite to remove an effect of illumination, wherein each pixel of each image in the series of images encodes a color value for a plurality of spectral bands, and wherein the illumination level comprises a per-spectral band illumination level; determine that a portion of the one of the series of images is part of a cloud; and generate a prediction of what was occluded by the cloud in the portion of the one of the series of images based in part on the modified color values.

9 . The computer-readable storage medium of claim 8 wherein each per-spectral band illumination level of an individual cloud-free image is computed by averaging the color values of that spectral band of the individual cloud-free image.

10 . The computer-readable storage medium of claim 8 , wherein each per-spectral band illumination level of an individual partially-cloudy image is computed based on a relative illumination, wherein the relative illumination of an individual image of the series of images is computed by dividing an average illumination of cloud-free pixels in the individual image by an average illumination of cloud-free pixels in a most cloud-free image of the series of images.

11 . The computer-readable storage medium of claim 10 , wherein pixels at a particular coordinate are included in the relative illumination computation when neither image has a cloud at the particular coordinate.

12 . The computer-readable storage medium of claim 8 , wherein each per-spectral band illumination level of an individual cloudy image is interpolated from adjacent images in the series of images.

13 . The computer-readable storage medium of claim 12 , wherein each per-spectral band of illumination is inferred by minimizing a difference between inferred illumination and observed illumination and limiting a rate of change of illumination between images in the series of images.

14 . The computer-readable storage medium of claim 8 , wherein the effect of illumination is removed from the color values of the one of the series of images captured by the satellite by dividing color values of the one of the series of images by a corresponding illumination level.

15 . A computing device, comprising:
at least one processor; and a computer-readable storage medium having computer-executable instructions stored thereupon which, when executed by the at least one processor, cause the computing device to:
determine an illumination level for each of a series of images captured by a satellite based on a cloud-weighted average of illumination values;
modify one or more color values of one of the series of images captured by the satellite to remove an effect of illumination;
determine that a portion of the one of the series of images is part of a cloud; and
generate a prediction of what was occluded by the cloud in the portion of the one of the series of images based in part on the modified color values and based in part on availability scores that estimate a likelihood that each pixel of each of the plurality of images is cloud-free.

16 . The computing device of claim 15 , wherein a cloud-weighted average of illumination values are computed by multiplying a relative illumination of an image of the series of images by a base illumination.

17 . The computing device of claim 15 , wherein the availability scores are multiplied by a difference between a predicted color value and an observed color value when iteratively minimizing a function that yields predicted versions of the series of images.

18 . The computing device of claim 15 , wherein relative illumination values of each of the series of images are computed using availability scores that indicate how cloud-free a pixel is.

19 . The computing device of claim 15 , wherein the prediction of what was occluded by the cloud is based on a predicted image that is based on multiplying, for each pixel, an availability score of that pixel by a difference between a predicted color value and a measured color value for that pixel.

20 . The computing device of claim 19 , wherein the prediction of what was occluded by the cloud is additionally made by minimizing changes in predicted color values of images that are adjacent in the series of images.

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Olsen Peder Andreas
• De Moura Estevao Filho Roberto
• Nunes Leonardo De Oliveira

Classifications

• G06V10/26

  Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion · CPC title

• G06V20/13

  Satellite images · CPC title

• G06V10/761

  Proximity, similarity or dissimilarity measures · CPC title

• G06V10/56

  relating to colour · CPC title

• G06T2207/10024

  Color image · CPC title