1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Scanners: Distortion between vertical and horizontal

Discussion in 'General Equipment Chat & Advice' started by neseho, Mar 9, 2021.

  1. neseho

    neseho Member

    Hello everyone,

    I have an Epson V500, and I've noticed that photographs appear sharply different when I scan them vertically (i.e. following the length of the scanner) or horizontally. They are narrower and stretched, or wider and shorter.

    Is this likely a fault with the scanner? Is there a gold standard way to scan photographs to ensure their dimensions aren't skewed?

    Thanks for any and all advice!
  2. Zou

    Zou Well-Known Member

    Have you checked that you aren't using a fixed output size? If you specify just the scan resolution without an output size it *should* retain original aspect ratio.
    Benchista likes this.
  3. Benchista

    Benchista Which Tyler

    Yup, sounds like a mismatch between input and output size ratios to me.
  4. PeteRob

    PeteRob Well-Known Member

    I've only ever scanned manually - i.e. done a pre-scan then selected the scan area. The area selected relates to the layout of the platten and not what is being copied. I do also see on turning on VueScan that there are options for both input size and output size as mentioned above. So it seems the obvious answer that the scanner is doing a requested scaling from one size to another. I've not used my (newish) Epson to scan prints yet only negatives in the holders. Epson Scan 2 automatically finds the negative frames, and mostly gets this correct, but the orientation is always fixed.
  5. neseho

    neseho Member

    Thanks for the responses. The target size I pick is always custom -- I drag and select the print shown in the preview scan.

    Items in the final scan appear elongated if scanned in one orientation compared with the other.
  6. PeteRob

    PeteRob Well-Known Member

    Odd. That shouldn't happen. What scanner software are you using ?
  7. neseho

    neseho Member

    I'm using Epson Scan, the bundled software.

    To give you a sense of the skew, I've just scanned the same photo in the two orientations. When it is vertical, the photo is 2968x4112 pixels. When horizontal, 2981x4088.
  8. neseho

    neseho Member

    I should add: While the difference there sounds slight, unfortunately it has a significant effect on things like faces and items,which look noticeably wider and shorter when compared side by side.
  9. PeteRob

    PeteRob Well-Known Member

    The scanner should indeed stick to its linear ppi setting.

    If you asked for 300 ppi then that's what you should get. I looked through the Epson support pages but didn't find anything similar. Make a square target of near platten-width and see what the scanned dimensions equate to. If they are out (i.e. a 8"x 8" target doesn't produce 2400 x 2400 pixels at 300 ppi then contact Epson support. If the scanner is well used it might be in need of a service.

    I said make it big and square because, if judging by pixel counting, if the original you are using is a small non-square picture your selection of scan area introduces errors in pixel count. It shouldn't lead to differences in the appearance of the scan made in photo mode.

    You can also try VueScan - it is free to try - and is one of the best scanner software around. It won't correct a scanner fault but it gives a chance to spot if some option selection is awry.

    Skew correction is for documents placed not precisely square on the platten. If you have it turned on then, I don't know how, this might be doing something. In photo mode as I understand it you should get a copy of whatever is under your selected area and if it isn't straight you have to fix it by rotating the result.
  10. neseho

    neseho Member

    Many thanks for your reply, Pete and others, I really appreciate your help with this.

    I couldn't find anything perfectly square, so instead I scanned two precise rulers perpendicular to each other at the same time. I found that with the 140mm ruler lying down the length of the scanner, the resulting picture was 1mm short of the scan taken width-wise. This 'error' is similar to the error I've detected in photos -- scanner lengths are 0.4-0.7% more compacted than identical scanned widths.

    Given that the width-wise scan is 'fixed' by the width of the sensor/mirror, I suspect the motor driving the equipment is not tracking distance correctly, and should be moving the scanning equipment ever so slightly faster than it is. This slowdown produces scans which are always 'squashed' in the length-wise dimension.

    Is this something which can likely be calibrated/resolved, or am I likely better off getting a new V600? :(
  11. PeteRob

    PeteRob Well-Known Member

    Ask Epson support. Could be that a drive belt is worn. It’ll come down to cost of repair vs replace
    Zou likes this.
  12. neseho

    neseho Member

    So I just tested this on my friend's V600 which is only 2 or 3 weeks old. Unfortunately, it has precisely the same issue. :(

    Would someone with a similar Epson model kindly do a test for me? Could they scan the same image in both orientations, rotate one, and then flick between the two to see if there is a change to the aspect ratio? I'm still getting this height/width issue.

    Someone on Reddit seemed to have the same issue a year or two back, he's uploaded photos which demonstrate the issue very well.

  13. PeteRob

    PeteRob Well-Known Member

    That at least suggests it is less likely a hardware issue. Try vuescan software. It is free to trial but isn’t expensive. Setting it up you might find a “switch” that is getting overlooked with the Epson software.

    I’ve got my (new) Epson scanner set up for negatives and I had a heck of a job connecting the head unit so I don’t want to unplug it until I’ve finished the negs (30 years worth).
  14. neseho

    neseho Member

    Hi Pete. I've tried this on Epson Scan, VueScan and Silverfast. Unfortunately, the issue exists on all three. I have no way of telling if the problem extends to negative scans as I can't scan the 35mm in a different orientation. The issue seems to relate to how distance is tracked by the reflective scanner.

    Rather alarmingly, this old thread on DP Review suggests that the error is non-linear, i.e. it is squashing the image at differing levels at different points:

  15. Andrew Flannigan

    Andrew Flannigan Well-Known Member

    The above problems are among the reasons why I generally "scan" negatives and transparencies with a camera...

    Bowens Lumitran with Canon 5 mounted 9521.jpg
  16. PeteRob

    PeteRob Well-Known Member

    Sounds remarkably similar.

    I just out of interest scanned a commercial 6x4 portrait print on my HP office 3:1 with Vuescan. It didn't look too bad but my choice of select was out. I had 1194x1776 px putting it "right way up" and 1794 x 1172 px putting it sideways on. Even though I was quite careful putting it on the platen and making my selection in vuescan I got quite different pixel dimensions. I didn't see any obvious distortion but it wasn't equally straight in the box. I'll try my Epson when I change it over to reflective scanning but it is a V850 Pro not a V500. I'd still try Epson support. They might have a calibration procedure. I know you can colour calibrate scanners I should think you could calibrate the stepping motor movement too.

    All I can suggest for the moment is to use the "best" orientation that minimises the distortion when scanning faces. Possibly, as I think one of the comments in the thread you linked to said, if the distortion is consistent you can correct in Photoshop. I think you can automate certain commands to make it easy and repeatable.
  17. Malcolm_Stewart

    Malcolm_Stewart Well-Known Member

    I've run a check on my Epson 4870 PHOTO of the same printed photo & scanned at 600dpi. The original orientation is horizontal.
    Scanned horizontally, the image is 4174 x 3012 pixels; scanned vertically, the image is 2993 x 4194. Rotating the second image by 90 degrees gives me 4194 x 2993. Ratio-wise the difference is 0.4%, despite my taking care to get the lasso exactly on the edges of the print in both orientations. Pixels are 12,572,088 and 12,552,642 resp. i.e. a ratio of 0.998 or 1.001. It would be nice if it was closer, but I imagine that the difference is something to do with stepper motors, and their step size.
  18. PeteRob

    PeteRob Well-Known Member

    Yes that would seem to be so but there is an implied spatial resolution that is quite high, e.g I’ve been scanning 35 negs at 2400 ppi to give images suitable for 12x8 prints at 300 ppi so in principle the stepper motors should resolve that, unless they move in bigger steps, scan in multiline strips and do some interleaving. Scanning at 300 or 600 ppi should be a doddle resolution wise.

    I wonder if leaving a margin around the photo, keeping the selection box unaltered between orientations, and scanning makes a difference, while paying attention to the input and output settings in case, somehow, the software is trying to force an aspect ratio on the selection and inserting extra scanlines that give the stretch effect.

    It’s all very strange.
  19. swanseadave

    swanseadave Well-Known Member

    I have an old Minolta Scan Dual and have no problem with vertical nor horizontal scans using Vuescan.
  20. PeteRob

    PeteRob Well-Known Member

    It seems to be specific to some Epson models - question is whether it is a real hardware fault or some obscure finger trouble.

Share This Page