Noise Reduction in Scanned Images Helmut Dersch (der@fh-furtwangen.de) Technical University Furtwangen September 14, 1999

PTAverager is a small utility which reduces noise in scanned images, and increases density D. Often, the dynamic range of scanners is limited by random noise of the CCD or the lamp. In that case averaging of several scans can be used to remove noise and improve image quality. This is especially useful in panoramic imaging where large dynamic ranges are common.

Installation:
PTAverager needs the Panorama Tools library 'pano12.lib' which can be downloaded from my website. It is contained in the packages 'Panorama Tools' and 'PTStitcher'. Drop the library on the System folder. Your system should move it to the extensions folder. Copy PTAverager wherever you like.

Operation:
Generate several identical scans of the same image. Be sure to scan with exactly the same settings, and do not touch the film holder. Any misalignment of the images will blur the final result. The more scans you make, the better will be the noise reduction. Using 2 images increases density D by 0.15, 5 images by 0.36 and 10 images by 0.5. Of course, at some point the random noise is suppressed below the detection limits of the device in which case no further improvement is possible.

Save all images in lossless PICT format. Then select all images from the finder, and drop them at once onto PTAveragers icon. You will be asked for a resultfile name. Then PTAverager reads one file after the other, and creates the final image.

Example:
The following image is underexposed. It is part of a slide scanned with the CanoScan 2700F at 1360dpi. There are some features in the shadows visible in the original, but they do not show in the scan. Adjusting levels in Photoshop and brightening the dark regions makes these features visible also, but severe noise becomes visible as well. This is shown in the right image, which is an enlargement of the white rectangle from the original.

To reduce noise 10 images where scanned with identical settings and averaged using PTAverager. While the averaged image looks identical to the left image above, the noise in the shadows is almost gone. Applying the exact same level adjustment and brightening leads to the following result

Notice that while noise is reduced, the sharpness of the original is preserved, ie no blurring occurred.