System for detecting structured artifacts in video sequences
Methods for detecting structured artifacts in a video may begin by accepting a video frame at a video input and then generating a filtered frame from the video frame. After a differenced image is produced, groups of connected pixels (blobs) within the differenced frame are identified, and isolated blobs are removed to produce a cleaned image. A Fourier transform is performed on the cleaned image, and structured artifacts may be identified based on a comparison of magnitudes of the Fourier components of the cleaned image. Devices that detect and identify structured artifacts are also described.
1. A method for detecting structured artifacts in a video, comprising:
accepting a video frame at a video input;
generating a filtered frame from the video frame;
differencing the filtered frame from the accepted frame to create a differenced frame;
thresholding the differenced frame;
comparing the thresholded differenced frame to a threshold minimum value of lit pixels in the frame to determine if the video frame contains a structured artifact;
identifying groups of connected pixels within the differenced frame as blobs;
removing isolated blobs to produce a cleaned image;
comparing a number of lit pixels in the cleaned image to the threshold minimum value of lit pixels;
when the number of lit pixels in the cleaned image is greater than the threshold minimum value of lit pixels, performing Fourier transform on the cleaned image; and
identifying the video frame as having structured artifacts based on a comparison of magnitudes of Fourier components of the cleaned image.
2. The method for detecting structured artifacts in a video according to claim 1 , further comprising:
marking regions of the video frame where the structured artifact occurs.
3. The method for detecting structured artifacts in a video according to claim 2 , in which marking regions of the video frame where the structured artifact occurs comprises:
thresholding the differenced frame to produce a binary image;
dilating the binary image;
identify blobs in the dilated binary image; and
generating bounding boxes.
4. The method for detecting structured artifacts in a video according to claim 2 , in which marking regions of the video frame where the structured artifact occurs comprises:
thresholding the differenced frame to produce a binary image;
dilating the binary image;
identify blobs in the dilated binary image; and
generating a centroid.
5. The method for detecting structured artifacts in a video according to claim 1 , in which removing isolated blobs to produce a cleaned image comprises, when there are no blobs within two-pixel vicinity, and when the blob has an area less than four pixels, replacing each pixel in the blob with a 0.
6. A video artifact detecting device for detecting structured artifacts in a video, comprising:
a video input for accepting a video frame;
a filter;
a difference generator structured to generate a difference image from the video frame and the video frame that has been filtered through the filter;
a blob identifier structured to identify groups of connected pixels within the differenced frame as blobs;
a thresholder for generating a thresholded output of the differenced frame;
a blob processor structured to compare the thresholded output of the differenced frame to a threshold minimum value of lit pixels in the frame;
a blob cleaner structured to remove isolated blobs to produce a cleaned image, when the number of lit pixels in the differenced frame is greater than the threshold minimum value of lit pixels, the blob processor further structured to compare a number of lit pixels in the cleaned image to the threshold minimum value of lit pixels;
a Fourier transformer for producing an output from the cleaned image, when the number of lit pixels in the cleaned image is greater than the threshold minimum value of lit pixels; and
an identifier structured to detect the video frame as having structured artifacts based on a comparison of magnitudes of Fourier components of the cleaned image.
7. The video artifact detecting device of claim 6 , in which the blob cleaner is structured to, when there are no blobs within two-pixel vicinity, and when the blob has an area less than four pixels, replace each pixel in the blob with a 0.
8. The video artifact detecting device of claim 6 , further comprising:
an artifact marker structured to generate locations of one or more artifacts.
9. The video artifact detecting device of claim 8 , in which the artifact marker comprises a binary image dilator.
10. The video artifact detecting device of claim 9 , in which the artifact marker further comprises a centroid generator.
11. The video artifact detecting device of claim 9 , in which the artifact marker further comprises a bounding box generator.
12. The video artifact detecting device of claim 6 , in which the blob cleaner is structured to, when there are no blobs within two-pixel vicinity on each side, and when the blob has an area less than four pixels, replace each pixel in the blob with a 0.