Anomalous pixel detection
View Patent ↗A video noise analyzer for detecting residual point noise in a video generates a list of candidate defective pixels by joining results of a temporal invariance detector and a spatial outlier detector. A data store is structured to contain data describing the candidate defective pixels and/or the blocks from which the pixels were selected. In other embodiments, a video noise analyzer includes a first salient analyzer structured to compare pixels within a block to a mean value of other pixels within the same block, a first threshold processor to generate a first pixel candidate list including pixels that differ significantly from their neighboring pixels, a second salient analyzer structured to generate a difference value between a subject pixel, its neighboring pixels, and neighbors of the neighboring pixels, and an identifier to label as a candidate dead pixel those pixels having a frequency occurrence of visual saliency greater than a predefined frequency threshold.
1. A video noise analyzer for detecting residual point noise in a video, comprising:
an input for receiving a video to be analyzed;
a temporal invariance detector structured to perform a differencing function between pixels of a same block of two adjacent frames and to generate a first pixel candidate list including pixels that have a relatively low rate of change between frames;
a spatial outlier detector structured to generate a second pixel candidate list including pixels that have a relatively high level of intensity deviation from adjacent pixels;
a combiner to combine the first pixel candidate list and the second pixel candidate list to generate a candidate noise pixel list; and
a data store structured to contain data describing one or more pixels on the candidate noise pixel list or the blocks from which the pixels on the candidate noise pixel list were selected.
2. The video noise analyzer for detecting residual point noise in a video according to claim 1 , in which the temporal invariance detector is used when the two adjacent frames span a scene change boundary.
3. The video noise analyzer for detecting residual point noise in a video according to claim 1 , in which the spatial outlier detector uses Root Means Squared or a median filter analysis.
4. The video noise analyzer for detecting residual point noise in a video according to claim 1 , in which the data store includes:
a block intensity window measure;
a block severity measure; and
a block detection confidence measure.
5. The video noise analyzer for detecting residual point noise in a video according to claim 1 , further comprising a color splitter to generate a red component frame of pixel data, a green component frame of pixel data, and a blue component frame of pixel data for each frame input to the video noise analyzer.
6. A method of detecting residual point noise in a video, the method comprising:
receiving a video to be analyzed;
performing a temporal invariance detection by performing a differencing function between pixels of a same block of two adjacent frames;
generating a first pixel candidate list including pixels that have a relatively low rate of change between frames based on the temporal invariance detection
performing a spatial outlier detection by selecting pixels that have a relatively high level of intensity deviation from adjacent pixels;
generating a second pixel candidate list including the pixels selected with a relatively high level of intensity deviation from adjacent pixels;
combining the first pixel candidate list and the second pixel candidate list to generate a candidate noise pixel list; and
generating a data store structured to contain data describing one or more pixels in the candidate noise pixel list or the blocks from which the pixels in the candidate noise pixel list were selected.
7. The method of detecting residual point noise in a video according to claim 6 , in which performing a temporal invariance detection occurs at a scene change.
8. The method of detecting residual point noise in a video according to claim 6 , in which performing a spatial outlier detection comprises performing a Root Means Squared function or a median filter analysis.
9. The method of detecting residual point noise in a video according to claim 6 , in which generating a data store data store includes:
generating a block intensity window measure;
generating a block severity measure; and
generating a block detection confidence measure.
10. The method of detecting residual point noise in a video according to claim 6 , further comprising, for each video frame input to the video noise analyzer, generating a red component frame of pixel data, generating a green component frame of pixel data, and a generating a blue component frame of pixel data.
11. A video noise analyzer for detecting dead pixels in a device by examining video made from the device, the video noise analyzer comprising:
an input for receiving a video to be analyzed;
a frame divider to generate a number of blocks of pixels from frames of the input video;
a first salient analyzer structured to compare pixels within a block to a mean value of other pixels within the same block;
a first threshold processor to generate a first pixel candidate list including pixels that differ significantly from their neighboring pixels;
a second salient analyzer structured to, for members of the first pixel candidate list, generate a difference value between a subject pixel, its neighboring pixels, and neighbors of the neighboring pixels; and
an identifier to label as a candidate dead pixel those pixels having a frequency occurrence of visual saliency greater than a predefined frequency threshold.
12. The video noise analyzer for detecting dead pixels according to claim 11 , in which the first salient analyzer is structured to compare pixels within a block only when a standard deviation of a block of pixels is below a predefined heterogeneous threshold.
13. The video noise analyzer for detecting dead pixels according to claim 11 , in which the first salient analyzer is structured to operate on combined color channels.
14. The video noise analyzer for detecting dead pixels according to claim 11 , in which the labels of candidate dead pixels is reset at a scene change.
15. The video noise analyzer for detecting dead pixels according to claim 11 , in which the neighboring pixels includes eight, ten, or twelve pixels.
16. A method for detecting dead pixels in a device by examining video made from the device, the method comprising:
receiving a video to be analyzed;
dividing frames of the received video into blocks of pixels;
performing a first salient analysis to compare pixels within a block to a mean value of other pixels within the same block;
generating a first pixel candidate list including pixels that differ significantly from their neighboring pixels;
performing a second salient analysis on members of the first pixel candidate list;
generating, based on the second salient analysis, a difference value between a subject pixel, its neighboring pixels, and neighbors of the neighboring pixels; and
identifying as a candidate dead pixel those pixels having a frequency occurrence of visual saliency greater than a predefined frequency threshold.
17. The method for detecting dead pixels according to claim 16 , in which performing a first salient analysis occurs only when a standard deviation of a block of pixels is below a predefined heterogeneous threshold.
18. The method for detecting dead pixels according to claim 16 , in which performing a first salient analysis comprises performing a first salient analysis on combined color channels.
19. The method for detecting dead pixels according to claim 16 , further comprising resetting a list of candidate dead pixels at a scene change.
20. The method for detecting dead pixels according to claim 16 , in which the second salient analysis is performed on eight, ten, or twelve pixels.