What's new

Removal of half-tone pattern from scan of newspaper


efiste2

Active Member
Messages
39
Likes
2
have any of you experts got any idea how to correct this, its a scan of an old newspaper cutting, but has a spotty effect, until you zoom in and it the spots disappear....is there a filter or effect i can use to correct it...unfortunately i didnt scan it, .:question:
H & Band Playing Main Street Paper Cutting.jpg
 

Eggy

Retired Moderator
Messages
5,312
Likes
5,080
I had a go at it:

1. Filter - Apply Gaussian blur 1 px
2. Filter - Apply Reduce noise 'strength' 10, 'preserve details' 100 %, 'reduce color noise' 100 %, sharpen details 0 %. Check the Remove JPEG Artifact box.

I came up with this:

test 02.jpg

Not perfect but the nasty half-tone is barely visible.
 

thebestcpu

Guru
Messages
1,349
Likes
1,168
There are some techniques within PS that can be used and some not available from within PS (or to the public yet).

First some info on best B&W image reconstruction of half tone newspaper print image that uses vary size dots.

Note that there is more information in the half tone image then most scans are set up to capture. When viewing half tone image using varying dot sizes, the human eye if viewing from a sufficiently large distance will integrate the “areas” of black with the “areas” of white to create the perception of a gray scale. To accurately reconstruct this same tone, one needs to have a scan where the areas are measured. Thought his can be considered overkill, scanning where the smallest pixel size of the scan is 1/2 the size to he smallest dot in the half tone image. From a practical distance the pile size should be about 1/10 the size of the distance between the dot centers (you can get by with 1/5 through not quite as good).

This is different from using an FFT. FFT works best for repeated identical patterns such as the regular stipple pattern such one can find in some matte prints or bad scans with patterns causing Moiré. FFT could be used in addition to the above technique and possible provide an even better result.

The scan should be done in grayscale or RGB (not bitmap), TIFF (not JPEG), and at the above mentioned resolution.

Without going into the math, then:

- Turn the gray tone/ RGB image into a Smart Object
- Apply the box blur filter and then slowly increase the input number from the minimum value of 1 until the pattern has just been eliminate (any higher and you just increase the blur.

Note that the above technique does not increase the resolution that is lost during half tonne, just more accurately recreates the gray tone levels and eliminating the annoying half tone pattern.

The GIF below is an example of the above workflow applied to a old newspaper article about my Great Grandmother. I only am showing the lamp on the table in the overall image since it had some details. It is a ceramic ball with some leaves and a rose on it (I have seen it in person ages ago). I was quite surprised at how much detail (or perceived detail) that could be recovered using this approach:

Click on image if you do not see animation.

Newspaper-Print-Recovery.gif


Now if you have a bitmap or JPEG at much lower resolution, you can’t get as good a recovery yet you can still try out the same technique.

In the OPs image, I upsized by 500% (my machine did not have enough memory to go to 1000% or 10X) using the “Nearest Neighbor” interpolation algorithm when resampling (important!). Then you apply the box blur image.

To demonstrate this approach, I have zoomed in to the Ops image to a chimney in the far background marked by the blue arrow in this image

EnhanceIT-SFW-Total.jpg

To show the the impacts of this approach I also included a GIF animation below cycling among the original image, eggy’s approach (very good) and the approach I documented above (maybe slightly better).

Click on image if you do not see animation.

EnhanceIT-PS-Options.gif


———————————————————————

Now that was the inside PS approach.

There are also some approaches that are very compute intensive and will probably not be adopted by Adobe given the average computers limitations.

Though not available to the public (yet), there were some NSA image algorithms being made available (indirectly) to research Universities. First stage is sending in the image for processing and the intermediate passes are provided back to the University. Full algorithm disclosure is to be released select Universities in the future. This program is headed at the NSA by Bob Smith and the technique is called Super Pass Object Oriented Fourier Transform (BS-SPOOFT). (I suspect this is some 3D version of the fast Fourier transform but what the heck do I know)

My son is a recently graduated PhD student at Rice University (Houston) and brought these NSA algorithms to my attention. Though I could not get direct access, I forwarded the OPs image and a very small section was processed through the multiple pass BS-SPOOFT.

I took the passes and overlaid them on the original image as a GIF sequence as supplemental frames in the prior GIF animation.

Note that this chimney “is” the chimney in the original image.

Click on image if you do not see animation.

EnhanceIT-with-BS-SPOOFT.gif


What do you think of that? I think it is incredible and almost seems too good to be true.
Well it was. What I really did was locate the exact location of the original image in google maps with the position on the street the same as the angle of the original image. It's a specific spot in Frodsham UK pointing almost due east. Here is a link to a zoomed in view of the chimney in Frodsham, UK (http://tinyurl.com/oymg2fy)I screen shot the picture and found a spot on the google maps shot that was unencumbered by the growth of trees and changes over decades of time and one of the chimneys was a perfect match (Told you it was the same chimney as in the OPs image. The rest is PS magic. Say, what can I say. Bob Smith was simply “BS” and the algorithm was just spoofed (“SPOOFT”). Sorry about that, yet the "in PS" approach is valid and I use it all the time. Everything else was just for grins. Thought it was time to add a little humor and hope no one takes offense for the short side diversion. You just never know who you can trust do do you?
 

MrToM

Guru
Messages
3,504
Likes
3,224
For no other reason than I just wanted to have a go...

newspaper_MT_01.gif

500x500px GIF dead centre.

Regards.
MrToM.
 

Eggy

Retired Moderator
Messages
5,312
Likes
5,080
thebestcpu:

Well, as a prank, this one counts!

I followed your message pic by pic and at the last pic one I was thinking 'what the f...'.

Hats off for the elaboration of your prank.

:thumbsup:
 

Tom Mann

Guru
Messages
7,223
Likes
4,340
thebestcpu - That was really funny!!! You should bring it out again on next April's Fools day. Just -- PLEASE -- don't spring it on any unsuspecting newbies. They will start breaking down the door with requests thinking we can do what the TV crime shows do, LOL.
Eggy - I like your method. Short and sweet.

I didn't have the time to try the FFT method on it, but from previous experience (e.g., https://www.photoshopgurus.com/forum...post1533699612) , my guess is that the results would look about the same as John's and Eggy's, but it takes quite a bit of manual effort to remove all the spikes in the Fourier plane.

I've also had pretty good success using a technique where I up-rez, make a few copies, displace each by part of one of the original pixels, average them all together, and down-rez as the last step. It produces results much like the up-rez, then box-blur method, but allows a bit more flexibility, IMHO.

Nice thread!

Tom M
 

thebestcpu

Guru
Messages
1,349
Likes
1,168
:cool2: Thanks Eggy and Tom Mann. Of course the joy of any prank is seeing/hearing the results. Thanks for responding to let me know. It made the elaboration worth the effort.
 

Top