Automatic Dust Removal In Digital Images

A processing device comprising:

• a processor;
• and
• a memory connected to the processor, the memory further comprising:
• instructions for estimating an attenuation of a plurality of pixels in a corrupted region of a digital image, the corrupted region including an artifact, instructions for applying an inverse transformation, based on the estimated attenuation, to ones of the plurality of pixels in the corrupted region,
• instructions for selecting a target patch area including at least some of the plurality of pixels of the corrupted region, the selected target patch area being based, at least in part, on having a highest per pixel average confidence level of all pixels included in the target patch area in comparison to a per pixel average confidence level of all pixels of each of a plurality of other target patch areas,
• instructions for determining a pixel distance between the selected target patch area and each of a plurality of candidate source patch areas outside of the corrupted region, such that ones of the plurality of pixels in the corrupted region that are included in the target patch area are considered when determining the distance,
• instructions for selecting at least one of the plurality of candidate source patches based on the determined distances,
• instructions for restoring ones of a plurality of pixels of the target patch area that are included in the corrupted region based on corresponding pixels of the selected at least one of the plurality of candidate source patches, and
• instructions for selecting a next target patch area, based, at least in part, on having a highest per pixel average confidence level of all pixels included in the target patch area in comparison to a per pixel average confidence level of all pixels of each remaining one of a plurality of other target patch areas.

A tangible machine-readable medium having recorded therein instructions which are executable by a processor, the instructions comprising:

• instructions for selecting a target patch area including at least some of a plurality of pixels of a corrupted region including an artifact, a plurality of pixels of the selected target patch area having a highest average per pixel confidence level from among a respective plurality of pixels included in each of a plurality of target patch areas,
• instructions for determining a pixel distance between the selected target patch area and each of a plurality of candidate source patch areas outside of the corrupted region, such that ones of a plurality of pixels in a portion of a corrupted region that is included in the target patch area, are considered when determining the pixel distance,
• instructions for selecting a source patch based on at least one of the plurality of candidate source patch areas having the determined pixel distance which is shorter than the determined pixel distances of others of the plurality of candidate source patch areas,
• instructions for estimating an attenuation caused by dust associated with the artifact,
• instructions for estimating a size of the dust,
• instructions for applying a median filter having a size proportional to the estimated size of the dust to produce an improved estimate of attenuation,
• instructions for applying an attenuation inversion, based on the improved estimate of attenuation, to recover damaged ones of the plurality of pixels in the selected target patch area,
• instructions for restoring the damaged ones of the plurality of pixels of the target patch area based on corresponding pixels of the selected source patch, and
• instructions for selecting a next target patch area, a plurality of pixels of the next target patch area based, at least in part, on a highest per pixel average confidence level in comparison to a per pixel average confidence level of a respective plurality of pixels of each of the plurality of target patch areas to be restored.