35mm Negative Scan line grain

Astrofiend

Hi everyone

Not sure where this should go.
I'm looking for some help with a scan I recently picked up. I'm using the Hp Scanjet g4010 to scan in 35mm negatives at 4800dpi and getting some strange grain patterns. I'm not sure if its to do with the film being out of date, the way it's been processed or the scanner itself.

Banding coloured grain seems to appear in whites of the photo in some shots its really noticeable others not so much.

Any tips of on how to reduce or fix this would be great, as noise reduction doesn't seem to affect it.

Examples:

i.imgur.com/5sLn6B3.jpg
i.imgur.com/TNXvdAb.jpg
i.imgur.com/oegt08Z.jpg

colourbox

This answer is not definitive, but...

The colored grain appears to be in near-white but not pure white areas. It also looks very digital in nature, the red spots don't look like film grain. Both observations suggest it's possibly digital noise from the scanner sensor. These are negatives, so very light areas are actually from the darkest part of the negative, and in digital the dark areas is where sensors are noisiest. Maybe it's a limitation of the sensor's dynamic range. Scanning enthusiasts look for the sensors with the widest dynamic range to get the lowest noise, but dynamic range is not easy to find out from spec sheets, and the scanners with great dynamic range for film are not cheap.

The horizontal lines look like a separate issue, possibly related to the precision of the scanner stepper motor. Check the HP scan software to see if it has multiple quality modes, try the highest quality/slowest speed mode.

I have a different scanner, a pure film scanner, and it can be set to do multipass scans to average out artifacts like scan lines and sensor noise, at the cost of a much longer time per scan. If the HP scan software has a higher quality option, it may include multipass scanning, or multipass might be a separate option that you have to turn on.

Astrofiend

> @colourbox said:
> This answer is not definitive, but...
>
> The colored grain appears to be in near-white but not pure white areas. It also looks very digital in nature, the red spots don't look like film grain. Both observations suggest it's possibly digital noise from the scanner sensor. These are negatives, so very light areas are actually from the darkest part of the negative, and in digital the dark areas is where sensors are noisiest. Maybe it's a limitation of the sensor's dynamic range. Scanning enthusiasts look for the sensors with the widest dynamic range to get the lowest noise, but dynamic range is not easy to find out from spec sheets, and the scanners with great dynamic range for film are not cheap.
>
> The horizontal lines look like a separate issue, possibly related to the precision of the scanner stepper motor. Check the HP scan software to see if it has multiple quality modes, try the highest quality/slowest speed mode.
>
> I have a different scanner, a pure film scanner, and it can be set to do multipass scans to average out artifacts like scan lines and sensor noise, at the cost of a much longer time per scan. If the HP scan software has a higher quality option, it may include multipass scanning, or multipass might be a separate option that you have to turn on.

Thanks very much, very informative. I've been playing with settings in vuescan, such as a slightly lower DPI and scanning raw TIFF negatives with no colour correction and inverting them in photoshop. This seems to reduce the grain lines as well as reduce the banding, its still not perfect but getting it to where I want it to be. I might try multiple passes to see if I can improve the scanner lines.

i.imgur.com/xAyoIfH.jpg

colourbox

Oh, nice then. I've used VueScan for years. If you go through the Input tab page in the Reference section of VueScan Help, you'll see a lot of the features I mentioned. Many are not revealed unless the Advanced option is on for the VueScan UI.

Astrofiend

If I do a multipass with Vuescan, does it automatically average out the scan to help with the lines, or do I have to do that manually in photoshop?

colourbox

Vuescan is going to save a single image file, so it will average its passes as it processes the one file it's going to save. The separate passes won't be provided as individual files to average in another program.

I just found the link confirming this:
https://www.hamrick.com/vuescan/html/vuesc29.htm#inputnumberofpasses

This option provides a similar function as the Number of samples option, but does not require the scanner to provide hardware support for multi-sampling. Each pass causes a full scan. After all passes are complete, the results are averaged, and the average is saved...On multiple passes the scanner head needs to be repositioned precisely at the same location as where it started on the previous pass. Some scanners do not support this operation. Others do, but reposition incorrectly, which will cause a blurred result.

It refers to the separate multi-sampling option, which should also be considered:

This option is available for scanners that support multi-sampling. As the scanner head passes over the media it makes multiple exposures for each location. The results for all samples are averaged.
This is a useful feature because any one exposure may be inaccurate, resulting in noise in the output. Noise will appear as one pixel whose color or tone is different than surrounding pixels. By taking multiple samples and averaging the results, the effect of inaccurate exposure is reduced.
This option will slow down scanning because the scanner is doing more. You should experiment with your scanner to see which balance of speed and accuracy is appropriate.
Number of samples is similar to Number of passes. Multi-sampling is preferable as the scanner head is positioned once, which ensures that the same area will be exposed for each sample. Multi-sampling is available only on a limited number of scanners.

It's important to remember the difference between multi-sampling and multi-pass. I have read all that before, but it had been a while and I obviously had to look it up again to keep the facts straight. My understanding of the above is:

• If the scanning hardware supports multi-sampling, use that first, because it does not carry the risk of bad alignment on multiple passes. Because the scanning head will stay in one place as it does multiple samples of the same line.
• If VueScan does not let you use multi-sampling on a particular scanner, that probably means the scanner can't do it, so you have to try multi-pass instead as the next best thing.

Hopefully you will find a combo that improves the results on your scanner!