Scan advice needed

[John Dolva]

Here is a scan with a portion enlarged by pixel resize (ie no interpolation). How do I adjust the scanning to minimise this noise which basically makes the scan useless?

There seems to be a relevant discussion here:

http://www.largeformatphotography.info/for...ead.php?t=11160

This bit seems particuarly interesting. However I can't understand it so I can use it.

Can someone explain please? How does it work?

"Multi-sample scanning will dramatically reduce noise. That is, adjusting the scanning software so that it samples each pixel repeatedly instead of just one time.

As a rough approximation, noise will fall as the square root of the number of samples - that is, if you sample each pixel four times, you'll have halved the noise level.

I regularly use 8x and 16x sampling on my Microtek 1800f and it produces dramatically superior results to 1x sampling.

If the Epson driver for the 4870 does not have the ability to do multi-sampling, switching to a second party scanner driver like Silverfast or VueScan will get you there.

The difference is dramatic. Don't give up on your scanner until you've given it a try."

[Frank Agbat]

Hi John,

I don't think the problem you're seeing is scanner-related, but is source-related. The process used to print the photograph in question created the problem, and scanning will not correct it.

This is why I keep harping about the need for a high-resolution repository of scans from the best available (preferrably original) sources...

The concept of multisampling is relatively simple. If you sample each pixel more than once and average the results, the effects of a mis-sample, bit-jitter, noise, etc. will be reduced. Of course, the penalty is that the time to scan is much longer. Also, if your native driver (the driver that came with your scanner) doesn't support multisampling, you'll have to find an aftermarket (third party) driver that will both work with your system and your hardware.

[John Dolva]

Hi John,

I don't think the problem you're seeing is scanner-related, but is source-related. The process used to print the photograph in question created the problem, and scanning will not correct it.

This is why I keep harping about the need for a high-resolution repository of scans from the best available (preferrably original) sources...

The concept of multisampling is relatively simple. If you sample each pixel more than once and average the results, the effects of a mis-sample, bit-jitter, noise, etc. will be reduced. Of course, the penalty is that the time to scan is much longer. Also, if your native driver (the driver that came with your scanner) doesn't support multisampling, you'll have to find an aftermarket (third party) driver that will both work with your system and your hardware.

Hi Frank, OK I'll look into that re multisampling and drivers etc.

re. Source::

I scanned an image, twice, once in portrait and once landscape. Here they are with values adusted to bring out the noise pattern.

(image)

This noise appears in a different pattern on the scans.

Also I compared it to another scan from a different book, and there the noise had far less of any pattern to it.

So I reason that it does indeed have to do with the source, but the way that the noise appears is somehow dependent on the scanning.

Any ideas?

Edited October 15, 2006 by John Dolva

[Jack White]

John...are you discussing scanning halftones?

My scanner program has a descreening feature which

works almost perfectly.

Jack

[John Dolva]

http://www.webopedia.com/TERM/H/halftone.html

"In printing, a continuous tone image, such as a photograph, that has been converted into a black-and-white image. Halftones are created through a process called dithering, in which the density and pattern of black and white dots are varied to simulate different shades of gray.

In conventional printing, halftones are created by photographing an image through a screen. The screen frequency, measured in lines per inch, determines how many dots are used to make each spot of gray. In theory, the higher the screen frequency (the more lines per inch), the more accurate the halftone will be."

.

http://www.scanhelp.com/288int/scontent/descreen.html

"What could be easier than just putting the document into the scanner, and then scanning it into a computer? Well many people don’t understand that the images that appear so clean and crisp in the magazine or newspapers are printed using a method known as halftoning. Those photos are printed using a series of overlapping dots that fool your eyes into seeing more colors than are actually there. Because of these overlapping dots, scanning these documents or images will produce something known as Moiré pattern (pronounced more-ay).

A Moiré pattern happens when two identical patterns of lines, circles, or array of dots are overlapped with imperfect alignment creating an interference pattern (resembles an optical illusion). Scanning these images into the computer will result in a wavy pattern that garbles the original image leaving you with a useless image. What makes things worse is scanning the image or document slightly off alignment can increase the effects of this wavy pattern, not only that, but the pitch of the scanners sensors will enhances the pattern even more. This will happen with any scanner, so don’t go blame it on the manufacturer just yet for the pitch of the scanner sensors."

___________________________

OK thank you for the keywords, Jack. I think I'm beginning to understand.

As I'm using someone elses scanner off site, I have to wait to see what the scanner has as far as adjustments. I suppose if I look through a loupe onto the page of the book I can see the halftone screen if it's there. Do all book b/w photo's use halftoning?

Another thing I'll try to set up to test is using a closeup lems on my digital camera to take a photo of the image and compare that. It'll take time to get around to that. But thank you all very much for this input. Any more much appreciated.

[Jack White]

John...all continuous tone images must be converted to DOTS to reproduce them.

On a computer the dots are called PIXELS.

On a printed page the dots are called HALFTONE DOTS.

If one attempts to RESHOOT either of such images, a MOIRE PATTERN will occur

where the screen patterns intersect. That is why an unwanted pattern appears

when you attempt to scan (pixels) an image from a book (halftone dots).

Remarkably a computer may be programmed to REVERSE THE HALFTONE

PROCESS almost perfectly within certain parameters. This is called

DESCREENING.

My scanner program is called SCAN WIZARD. It has such a feature (see

attachment) which I use frequently. Probably most scan programs have such

a feature.

Jack

PS...yes, all photos in books are HALFTONES, made up of tiny dots of varying

sizes. In BW, all the dots are BLACK; in color photos, the dots are separated

with filters in to YELLOW, CYAN and MAGENTA, plus usually a highlight scan in

BLACK to add detail. Color photos are thus said to be in FOUR-COLOR PROCESS.

A sort of exception is ROTOGRAVURE, a printing process which bleeds the

ink into the paper in a fuzzy manner in which the halftone dots are less obvious,

giving an apparent effect of a continuous tone. Rotogravure was popular in the

mid 1900s, mostly in movie fan magazines and such, since the results looked

more like actual photos than halftones.

Edited October 16, 2006 by Jack White

[John Dolva]

Thank you Jack, it sounds like descreening the halftone and some multisampling is the way for the best result?

image from: http://www.brightonart.org/freeframe/galle...s/halftone.html

this is a sample of an enlarged portion of a photo showing the halftone 'screen'. The halftone descreener would presumably remove the black and white dots.

Now this image can be further resized.

Here is a pixel resize (without any interpolation to 'bleed' the data), This shows the pixels enlarged::

(image)

It's not quite the same thing. One halftone 'dot' in this instance would be about a grid of 7x7 pixels?

Now this is my present understanding...if anyone can elaborate, correct, clarify please do so::

Pixel is more like a unit of measurement, each pixel is part of a data map of the image without space in between. The screen (monitor, LCD, CRT etc) displays the data with a finite screen 'pitch', similar to the halftone screen but not quite. In the case of the CRT there is a physical mesh or screen that is between the ray emittor and the glass 'screen', the number of these screen holes per inch gives the dots per inch. How many DPI displays one pixel?

Edited October 16, 2006 by John Dolva

[Jack White]

Thank you Jack, it sounds like descreening the halftone and some multisampling is the way for the best result?

image from: http://www.brightonart.org/freeframe/galle...s/halftone.html

this is a sample of an enlarged portion of a photo showing the halftone 'screen'. The halftone descreener would presumably remove the black and white dots.

Now this image can be further resized.

Here is a pixel resize (without any interpolation to 'bleed' the data), This shows the pixels enlarged::

(image)

It's not quite the same thing. One halftone 'dot' in this instance would be about a grid of 7x7 pixels?

Now this is my present understanding...if anyone can elaborate, correct, clarify please do so::

Pixel is more like a unit of measurement, each pixel is part of a data map of the image without space in between. The screen (monitor, LCD, CRT etc) displays the data with a finite screen 'pitch', similar to the halftone screen but not quite. In the case of the CRT there is a physical mesh or screen that is between the ray emittor and the glass 'screen', the number of these screen holes per inch gives the dots per inch. How many DPI displays one pixel?

John...Dots Per Inch in halftone vary between 55 dpi and 300 dpi. The

quality of paper to be used determines. Coarse newsprint ususally

requires 55, 65, or 85. Slick enamel paper ususally requires 100 up

to 133. Very fine printing in expensive books usually uses 133-300.

I understand DPI and Halftone Dots are not the same, or even the

same shape. I was using that only as a comparison.

Hope I have helped.

Jack

[Frank Agbat]

John,

Please keep in mind that computers can NOT recover data that is lost in the halftone process.

Once a photograph is halftoned, the data that was originally available is lost. The halftone process reduces, in effect, both the bit-depth AND the resolution of a photograph.

Descreening and other digital algorithms do a very good job of approximating what might have been in the original. To the eye, the resulting image is improved notably. However, the actual truth of the matter is that the results are an approximation based on the available halftone dots and the algorithm used.

So yes, various techniques (moire reduction, multisampling, descreening, smoothing, etc) can improve the visual appearance of an image dramatically to the eye, they cannot recover lost information. Haltoning is not a lossless compression technique. Quite the opposite -- it is tremendously lossy as the resolving power of the film (lines per millimeter) is often far greater than even hundreds of dots per inch in the printing process. As such, once a photograph has been half-toned, the only real usefulness of the resulting image is macroscopic examination.

So when Jack says, "Remarkably a computer may be programmed to REVERSE THE HALFTONE PROCESS almost perfectly within certain parameters. This is called DESCREENING.", I must respectfully disagree, at least to some extent. As I noted above, these algorithms are inevitably approximations. Within small areas, it is conceivably possible that you could get lucky and have results that appear to be perfect. However, when the images are compared digitally, the actual scan elements / pixels are different. If I get some time today, I'll try to post an example of what I'm talking about.

[Jack White]

John,

Please keep in mind that computers can NOT recover data that is lost in the halftone process.

Once a photograph is halftoned, the data that was originally available is lost. The halftone process reduces, in effect, both the bit-depth AND the resolution of a photograph.

Descreening and other digital algorithms do a very good job of approximating what might have been in the original. To the eye, the resulting image is improved notably. However, the actual truth of the matter is that the results are an approximation based on the available halftone dots and the algorithm used.

So yes, various techniques (moire reduction, multisampling, descreening, smoothing, etc) can improve the visual appearance of an image dramatically to the eye, they cannot recover lost information. Haltoning is not a lossless compression technique. Quite the opposite -- it is tremendously lossy as the resolving power of the film (lines per millimeter) is often far greater than even hundreds of dots per inch in the printing process. As such, once a photograph has been half-toned, the only real usefulness of the resulting image is macroscopic examination.

So when Jack says, "Remarkably a computer may be programmed to REVERSE THE HALFTONE PROCESS almost perfectly within certain parameters. This is called DESCREENING.", I must respectfully disagree, at least to some extent. As I noted above, these algorithms are inevitably approximations. Within small areas, it is conceivably possible that you could get lucky and have results that appear to be perfect. However, when the images are compared digitally, the actual scan elements / pixels are different. If I get some time today, I'll try to post an example of what I'm talking about.

I agree with what Frank says. I should have said "for practical purposes of

APPROXIMATING what the original shows". I said "almost perfectly within

certain parameters", which is CORRECT but not very specific.

Frank is technically correct that the information IS LOST in the SPACES BETWEEN

THE DOTS. However, the parameters I did not detail get rather technical,

having to do with DOTS PER INCH and what the unaided eye can perceive.

As an art director for fifty years, I ordered thousands of halftones made.

Parameters to consider begin with the paper to be used and the printing

process (lithograph or letterpress). Briefly, the better the paper and the

better the printing process, the greater quality of the halftone.

Letterpress newspaper of halftones commonly used 65-line screens...but

present day newspapers use lithography and 120-line screens.

Letterpress halftones on good paper commonly required 110-line screens,

but letterpress is now totally obsolete, with lithography accounting for

virtually all printing, and most lithograph screens now 133-line to 300-line.

So all halftones are not equal in quality loss. The finer the screen, the less

the loss.

AS FAR AS PERCEPTION BY THE UNAIDED HUMAN EYE IS CONCERNED,

A QUALITY HALFTONE OVER 133-LINE CAN BE COMPARED TO THE ORIGINAL

COPY AND THE UNAIDED EYE CANNOT PERCEIVE THE DIFFERENCE. In

press checks, I did this hundreds of times...comparing the press sheet

to original photos. A printer's loupe (8x) was used to examine the dots

for crispness...but the unaided eye could not see the dots.

But Frank is correct...some loss occurs in the halftone process, but it

is not always observable without magnification.

Jack

[Jack White]

John,

Please keep in mind that computers can NOT recover data that is lost in the halftone process.

Once a photograph is halftoned, the data that was originally available is lost. The halftone process reduces, in effect, both the bit-depth AND the resolution of a photograph.

Descreening and other digital algorithms do a very good job of approximating what might have been in the original. To the eye, the resulting image is improved notably. However, the actual truth of the matter is that the results are an approximation based on the available halftone dots and the algorithm used.

So yes, various techniques (moire reduction, multisampling, descreening, smoothing, etc) can improve the visual appearance of an image dramatically to the eye, they cannot recover lost information. Haltoning is not a lossless compression technique. Quite the opposite -- it is tremendously lossy as the resolving power of the film (lines per millimeter) is often far greater than even hundreds of dots per inch in the printing process. As such, once a photograph has been half-toned, the only real usefulness of the resulting image is macroscopic examination.

So when Jack says, "Remarkably a computer may be programmed to REVERSE THE HALFTONE PROCESS almost perfectly within certain parameters. This is called DESCREENING.", I must respectfully disagree, at least to some extent. As I noted above, these algorithms are inevitably approximations. Within small areas, it is conceivably possible that you could get lucky and have results that appear to be perfect. However, when the images are compared digitally, the actual scan elements / pixels are different. If I get some time today, I'll try to post an example of what I'm talking about.

I agree with what Frank says. I should have said "for practical purposes of

APPROXIMATING what the original shows". I said "almost perfectly within

certain parameters", which is CORRECT but not very specific.

Frank is technically correct that the information IS LOST in the SPACES BETWEEN

THE DOTS. However, the parameters I did not detail get rather technical,

having to do with DOTS PER INCH and what the unaided eye can perceive.

As an art director for fifty years, I ordered thousands of halftones made.

Parameters to consider begin with the paper to be used and the printing

process (lithograph or letterpress). Briefly, the better the paper and the

better the printing process, the greater quality of the halftone.

Letterpress newspaper of halftones commonly used 65-line screens...but

present day newspapers use lithography and 120-line screens.

Letterpress halftones on good paper commonly required 110-line screens,

but letterpress is now totally obsolete, with lithography accounting for

virtually all printing, and most lithograph screens now 133-line to 300-line.

So all halftones are not equal in quality loss. The finer the screen, the less

the loss.

AS FAR AS PERCEPTION BY THE UNAIDED HUMAN EYE IS CONCERNED,

A QUALITY HALFTONE OVER 133-LINE CAN BE COMPARED TO THE ORIGINAL

COPY AND THE UNAIDED EYE CANNOT PERCEIVE THE DIFFERENCE. In

press checks, I did this hundreds of times...comparing the press sheet

to original photos. A printer's loupe (8x) was used to examine the dots

for crispness...but the unaided eye could not see the dots.

But Frank is correct...some loss occurs in the halftone process, but it

is not always observable without magnification.

Jack

[Jack White]

Attached Mcintire photo descreened from a halftone. It compares favorably

with the printed halftone, but obviously is not as good as the continuous

tone original.

Jack

[Frank Agbat]

John,

Please keep in mind that computers can NOT recover data that is lost in the halftone process.

...

I agree with what Frank says. I should have said "for practical purposes of

APPROXIMATING what the original shows". I said "almost perfectly within

certain parameters", which is CORRECT but not very specific.

Frank is technically correct that the information IS LOST in the SPACES BETWEEN

THE DOTS. However, the parameters I did not detail get rather technical,

having to do with DOTS PER INCH and what the unaided eye can perceive.

As an art director for fifty years, I ordered thousands of halftones made.

Parameters to consider begin with the paper to be used and the printing

process (lithograph or letterpress). Briefly, the better the paper and the

better the printing process, the greater quality of the halftone.

Letterpress newspaper of halftones commonly used 65-line screens...but

present day newspapers use lithography and 120-line screens.

Letterpress halftones on good paper commonly required 110-line screens,

but letterpress is now totally obsolete, with lithography accounting for

virtually all printing, and most lithograph screens now 133-line to 300-line.

So all halftones are not equal in quality loss. The finer the screen, the less

the loss.

AS FAR AS PERCEPTION BY THE UNAIDED HUMAN EYE IS CONCERNED,

A QUALITY HALFTONE OVER 133-LINE CAN BE COMPARED TO THE ORIGINAL

COPY AND THE UNAIDED EYE CANNOT PERCEIVE THE DIFFERENCE. In

press checks, I did this hundreds of times...comparing the press sheet

to original photos. A printer's loupe (8x) was used to examine the dots

for crispness...but the unaided eye could not see the dots.

But Frank is correct...some loss occurs in the halftone process, but it

is not always observable without magnification.

Jack

Jack,

Well said and agreed. I also wanted to agree that some of the newer litho processes maintain color registration and numerous other parameters with amazing precision. The results can be quite impressive indeed.

The problem of halftone loss shows up best under magnification, as you pointed out. As such, when one attempts to enlarge a small portion of such an image, the results are exactly what John is experiencing. The small area, in effect, just doesn't have enough dots when halftoned to faithfully reproduce small features. With a descreen process during either scanning or image adjustment, the image may end up looking even better to the naked eye, but the small features are unreliable, as they have been approximated by software.

Your image shows this concept quite nicely. The image looks good to the eye and doesn't seem to have any moire pattern, at least not that I could see easily. While this is a good image, I would not expect to be able to determine if, for example, a gentleman waaaay off in the distance had forgotten to shave! Such tiny detail, even if the film/lens was able to resolve it, would probably not survive the halftone process.

What John needs, if he wants to analyze features in the 6th floor window, is a scan that doesn't use a halftoned image as its source (and preferrably doesn't have jpg compression artifacts, either).

Respectfully,

Frank

[John Dolva]

"unreliable, as they have been approximated by software."

yes, this is what I have found. A number of rescans showed precicely this. The halftoning can be reduced but the result, while looking better largescale, can't be trusted on the fine detail. It seems to me to be impossible to be sure beyond a certain degree what is original and what is created.

It's probably possible to work out some rule of thumb for this in terms of power of magnification. I wonder if some thought has been given to this? What would be the considerations for an 'average' book photo scan? Is this possible to say?

________________

Thank you very much Jack and Frank.

Good to know. It supports very much a call for a common source of a researcxh community certified 'best' copies. Perhaps a committee of sorts could be formed to work on that? Or at least consider laying the groundworks for thinking in assembling such a collection? Perhaps those assembling at the conference this year could give it some thought?

[Jack White]

Often researchers must use what is at hand, which is most often

a halftone. Few researchers have access to originals.

When doing this, researcher must be aware that halftones may

introduce artifacts not present in the original. Most prominent

false conclusions are the early studies of Raymond Marcus, who

thought he saw a man with a gun in a Moorman halftone,

below. The image does NOT appear in ANY continuous tone

Moorman copy.

Jack