PDA

View Full Version : Removing unwanted Texture from Digital Images



mongo
10-03-2011, 8:38am
To reproduce some old wedding photos for an anniversary shortly (for which there are now no negatives), Mongo has been photographing to old original photos to reprint them. Problem is that the original photos (about 40 - 50 years ago) were printed on fancy textured paper. So, does any one have any ideas about how Mongo can tone down (or even remove ) the textured look from the digital images before they are printed. Mongo only has photoshop CS3 and some noise reduction software to work with.

Any and all suggestion gratefully accepted (except the usual suggestion telling Mongo to get lost).

dbax
10-03-2011, 9:01am
Hey Mongo, my initial thoughts are that you will need to apply some type of blur to remove the texture, I guess I'd try blurring a new layer of the original and either adjusting the opacity or masking and painting back in the blur to smooth the texture. Interesting problem, I'll give it some more thought :D

dbax
10-03-2011, 9:20am
A quick search Mongo, this link may help http://www.retouchpro.com/forums/photo-restoration/8971-remove-photopaper-texture-old-photo.html

Analog6
10-03-2011, 9:37am
Try the Median Choice in Photoshops sharpen filter - I have used it to remove the screen overlay from photos scanned from newspapers with some success.

mongo
10-03-2011, 9:42am
very many thanks David. Mongo is not a member of that site but has the both tools they discussed and will give it a go from the description of the steps to take.

ameerat42
10-03-2011, 11:06am
I have found Neat Image to do a better job than the P'shop filters, but the free version ONLY works on images less than 1000 pixels on a side. So what I do is make up complementary portions of selection marquees and neatimage (new verb) each one at a time.

It's a problem getting a clean image in the first place from such a photo. I used to try lamps at all angles and re-photograph them in the old days, with little success. Only thing I didn't do then - and now I don't have to - was to use a light tent to photograph the original image. I'd say though, the likes of NeatImage may be the best help.
Am.

PS: And then apply some mild shrpening with Photoshop's Smart sharpen, not USM.

kiwi
10-03-2011, 11:31am
pm the member - damo77 mongo...he does that stuff for a living

mongo
10-03-2011, 11:50am
thanks guys for your additional ideas.

dbax
10-03-2011, 11:54am
I'd be interested to hear how you get on with this Mongo, maybe a before and after with a few step/tips ?

mongo
11-03-2011, 4:45pm
OD - Mongo is finding this a little frustrating and disappointing. Mongo is getting varying results using a combination of neat image and gaussin blur. The main variable is how good or bad Mongo re-took the photos. Maybe they were too accurately photographed using a very sharp macro lens and this unfortunately caught more of the unwanted details than Mongo would have hoped for. This is one case where less accurate could mean a better outcome for this exercise. Still trying a number of different things and will let you know. In the interim, here is what somebody suggested on that other site you pointed Mongo to but Mongo has not tried this yet..

"• I duplicated the Background Layer twice.
• On the first duplicate I run Neat Image (defaul settings).
• On the second duplicate (on top) I run Gaussian Blur, added some noise and adjusted its Opacity. I, then, added a Layer Mask and, with a soft black brush, I painted over eyes, nose and mouth to uncover those sharper features from the underlying 'Neat Image' Layer.
• Merged Visible.
• Created a Levels Adjustment Layer to improve contrast.
• Created a Hue/Saturation Adjustment Layer (colorize option checked) to give a light tint to the picture.
• Added a little Noise + Soft Gaussian Blur (radius = 0.3-5).
• Created a New empty Layer > Blending Overlay and with a soft black/white Brush (Opacity 10-20%) I painted over areas I wanted to emphisize/highlight.
• I further enhanced the contrast using Unsharp Mask (Amount>20%, Radius>50, Threshold>0).
• I finally sharpened the image using USM."

Damo77
11-03-2011, 5:13pm
http://retouchpro.com/tutorials/?m=show&id=185

ameerat42
11-03-2011, 5:27pm
Mongo. Don't pull your fur out! Have you got a sample we can look @ & work on?
Am.

mongo
11-03-2011, 7:46pm
Old Dog - a quick update. Damo77's specific suggestion led to software directly for this problem. Have downloaded it and it gives a reasonably good result by removing about three quarters of the unwanted effect. It is specifically a "fourier Transform Filter" which can be a plug-in for photoshop and some other similar softwares. Big thanks to Damo77 and Kiwi for putting Mongo onto him.

Ameerat42, very kind offer and Mongo will take you up on it after dinner tonight he will post a sample photo and see what you can do with it over the next few days if you wish. Again, Mongo's thanks for your offer. BTW, you are right about the fur - winter is coming after all. :D

mongo
11-03-2011, 9:52pm
Ameerat42 - as promised a sample to play with is attached - good luck . If anyone else feels like having a go and providing details - please do.

cheers
Mongo

FallingHorse
11-03-2011, 11:53pm
Oh man - that is some texture - good luck with it :eek:

ameerat42
13-03-2011, 2:22pm
Ta, Mongo. Give us a tick. Have been laid low for a day and a half. Just got back on line. Am.
Hmm. Sure is heavy texture...

Damo77
13-03-2011, 3:15pm
http://www.bellephotography.net.au/pics/R-&-M-done.jpg

ameerat42
13-03-2011, 9:59pm
Well, Mongo. I have had a go trying to do this using only the tools available in CS2 PS, with the exception of one hit of Neat Image. In case you can't see the PS actions in EXIF they are below.
A couple of times I repeated actions, since to apply them more strongly, or twice in succession, did not give pleasing results. It was a harder image to work on than I thought it would be. One thing I didn't do was extract any info from the dark areas. I even checked with Velvia Vision but there was not enough DR. (Maybe a 48-bit scan...) Finally, I didn't try to mask off and remove all the texture. There are residual amounts in the bg car upholstery and some just noticeable on the bride's dress. However, I thought those bits blended in with the image somewhat.

Unfortunately, I had to save it as jpeg quality 7 (medium). Hope the result doesn't scare you into the bush.
Am.

Surface Blur: 5, 15 - (1st major reduction of texture)
Neat Image: (Default); fade 66% (non-Adobe CS native action)
Add Noise: Gaussian, 1.5, Mono - (helps hide blemishes and gives a "surface" to sharpen
Dust/Scratches: 1, 0 - (reduce some of the more obvious white splotches)
Smart sharpen: 50%, 1, Lens Blur
Hue/Sat: Reds +5, -5, 0 - (try to make the colors less garish)
Dust/Scratches: 1, 0 - (as above for residuals)
Repeat Add Noise: Gaussian, 1.5, Mono
Repeat Smart sharpen: 50%, 1, Lens Blur
Auto contrast; fade 66% - (mildly enhance contrast)

mongo
13-03-2011, 10:05pm
thanks Damo77 - did you use the fourier Transform Filter software or that plus something else or just something else ?? It certainly has cleaned up the texture but lost a fair percentage of sharpness (which was not great to start with).

Mongo tried the fourier Transform Filter and as stated cleans up about 70% of the texture without too much loss of detail. It may be a matter for compromise between getting rid of some texture but holding onto as much detail as possible.

Mongo is very grateful for your suggestions and efforts on this problem.

Ameerat42 - equally big thanks for your large efforts also. The level and location of remaining texture is more than acceptable in your resulting sample. It does have a slightly "pasty" look with defined noise. Mongo will have to maybe blend techniques and compromise between one thing and the other as stated above. To get any improvement on this difficult image is more than anyone could hope for and we have achieved that . Again , a very big thanks to you also.

mongo
16-03-2011, 3:51pm
For those who helped and may be wondering, this is as good as Mongo could manage. Again thanks for your help and ideas.