Passive camera lens smudge detection
This disclosure relates to devices, systems, and methods for passively determining the probability of the presence of a camera lens smudge on an image capture device, e.g., based on an analysis of images captured by the image capture device (and without the need for any reference images). The techniques include: obtaining an image using an image capture device and identifying and analyzing (e.g., in the form of gradient analysis) the regions of interest (ROIs) in high intensity areas of the image. A probability of the presence of a smudge on the image capture devices may then be determined and outputted to a display using any desired modality. According to some aspects, machine learning (ML)-based models may be used in the performance of the passive camera lens smudge detection techniques disclosed herein. According to other aspects, temporal consistency and/or sensor consistency constraints may also be applied to the smudge detection processes.
1 . A device, comprising:
a display;
an image capture device; and
one or more processors operatively coupled to memory, wherein the one or more processors are configured to execute instructions causing the one or more processors to:
obtain an image using the image capture device;
identify a region of interest (ROI) in a high intensity area of the image;
determine an image gradient of the ROI is greater than a threshold value;
extract a plurality of sub-ROIs for the ROI;
determine a probability of a smudge on a lens of the image capture device; and
output the probability of the smudge on the lens of the image capture device to the display.
2 . The device of claim 1 , wherein the probability of the smudge on the image capture device is determined by a contrast measure for each of the plurality of sub-ROIs.
3 . The device of claim 1 , wherein the probability of the smudge on the image capture device is determined using convolutional blocks of a neural network for each of the plurality of sub-ROIs.
4 . The device of claim 1 , further comprising: obtaining a second image using the image capture device, wherein the probability of the smudge on the image capture device is further determined based, at least in part, on an analysis of high intensity areas of the second image.
5 . The device of claim 1 , wherein the probability is output in the form of a message on the display.
6 . The device of claim 1 , wherein the probability is output in the form of a status bar on the display.
7 . An image processing method, comprising:
obtaining an image using an image capture device;
identifying a region of interest (ROI) in a high intensity area of the image;
determining an image gradient of the ROI is greater than a threshold value;
extracting a plurality of sub-ROIs for the ROI;
determining a probability of a smudge on a lens of the image capture device; and
outputting the probability of the smudge on the lens of the image capture device to a display.
8 . The method of claim 7 , wherein the probability of the smudge on the image capture device is determined by a contrast measure for each of the plurality of sub-ROIs.
9 . The method of claim 7 , wherein the probability of the smudge on the image capture device is determined using convolutional blocks of a neural network for each of the plurality of sub-ROIs.
10 . The method of claim 7 , further comprising: obtaining a second image using the image capture device, wherein the probability of the smudge on the image capture device is further determined based, at least in part, on an analysis of high intensity areas of the second image.
11 . The method of claim 7 , wherein the probability is output in the form of a message on the display.
12 . The method of claim 7 , wherein the probability is output in the form of a status bar on the display.
13 . A system comprising:
an image capture device;
a display; and
a smudge detection module, wherein the smudge detection module is configured to:
obtain an image using the image capture device;
identify a region of interest (ROI) in a high intensity area of the image;
determine an image gradient of the ROI is greater than a threshold value;
extract a plurality of sub-ROIs for the ROI;
determine a probability of a smudge a lens of on the image capture device; and
output the probability of the smudge on the lens of the image capture device to the display.
14 . The system of claim 13 , wherein a neural engine comprises the smudge detection module.
15 . The system of claim 13 , wherein the probability of the smudge on the image capture device is determined by a contrast measure for each of the plurality of sub-ROIs.
16 . The system of claim 13 , wherein the probability of the smudge on the image capture device is determined using convolutional blocks of a neural network for each of the plurality of sub-ROIs.
17 . The system of claim 13 , further comprising: obtaining a second image using the image capture device, wherein the probability of the smudge on the image capture device is determined using the second image.
18 . The system of claim 13 , wherein the probability is output in the form of a message on the display.
19 . The system of claim 13 , wherein the probability is output in the form of a status bar on the display.
20 . The system of claim 13 , wherein the smudge detection module is operatively connected to an image signal processing (ISP) pipeline.