HP scientists have developed a technology that can detect creases on photos using a standard, unmodified flatbed scanner. Once the photo is properly scanned into the computer, the software can determine where the defect is in the photo and then manually correct it to remove any traces of creases or folds.
On the surface, this technology seems to be relatively simple in the printing market. Most flatbed scanners use two separate light bulbs to accurately capture all colors on a photo. The same picture can be captured by controlling these two slightly different images that are independent of each other (bulbs move in different directions under the photo). This can generate basic three-dimensional information.
As defects, such as creases, are identified, software can completely cover them artificially. This is called "padding" and each pixel in the scanned crease is replaced with a new pixel, which is taken very close to the crease. The software can ensure that two pixels are similar to avoid a huge red dot in the middle of a green pixel.
This allows you to scan a creased photo as a JPEG file without smearing, without requiring any retrofit hardware and expertise for the user.
Using existing (HP) scanners, this technology has proven to be feasible in the laboratory, but researchers have immediately pointed out the defects of the technology. Any defects in the plane that are close to the glass will not be found, the two bulbs in the scanner must be the same, and the software has problems with large creases or tears.
So far no commercial application of this technology has been announced, but interested parties can read detailed technical documentation.