Jump to content

is this Moiré pattern on some of my scans?


Recommended Posts

when I scan Kodak 5222 with my Minolta DImage Scan Dual IV, I see this pattern:

 

 

http://norayr.am/weblog/uploads/2018/10-28/2018-10-07-0037_sm.jpg

 

http://norayr.am/weblog/uploads/2018/10-28/2018-10-07-0038_sm.jpghttp://norayr.am/weblog/uploads/2018/10-28/2018-10-07-0041_sm.jpghttp://norayr.am/weblog/uploads/2018/10-27/2018-10-07-0035_sm.jpg

 

It also appeared with AGFA 100, but I usually don't get this problem.

 

Lets say this is Kodak Tri-X pushed to 1600 and it looks okay.

 

http://norayr.am/weblog/uploads/2018/10-44/2018-10-07-0026_sm.jpghttp://norayr.am/weblog/uploads/2018/10-42/2018-10-07-0009_sm.jpg

 

So is this moire pattern? Why does it appear on some films, and does not appear on others?

I scanned the same film with CanoScan 9000 mk ii and it's ok, no such pattern. Is something wrong with my Minolta scanner?

What can I do to avoid it?

Link to comment
Share on other sites

It could be Moiré if there was something there with high spatial frequencies.

 

I might imagine it on some, but not on the sky.

 

Well, there is something called grain aliasing, which is pretty much Moiré on the grain.

 

Normally scanners are designed such that the optical resolution is lower than the sensor resolution, such that it isn't a problem.

 

Non-interchangeable lens cameras also do that, but with interchangeable lenses, especially with an existing lens series, they can't do that.

-- glen

Link to comment
Share on other sites

Moire can only occur when sampling a regularly spaced pattern, and grain is never that regular.

 

I don't think the ScanDual IV supports IR dust removal, otherwise I'd suspect that.

 

If there's a glass plate in contact with the film it might be Newton's rings. There again the patterning doesn't look typical of those.

 

My best guess is scanner 'stuttering', where the scanner film transport doesn't run smoothly. Maybe it's time for a CLA?

 

I also once had a weird effect where I got mains-frequency crosstalk with the sensor signal. I was tinkering with a scanner with its cover and electrical shielding removed. The result was a bit like that. Try re-routing any power cables away from the scanner.

Link to comment
Share on other sites

It's not Moire.

 

I see lateral scan lines in your examples, which occur when you use the standard, multi-row scan option. There may be minor differences between the three rows of sensors. They go away if you use the fine scan option, which unfortunately takes three times as long. Lines are generally seen in areas of low contrast and medium density, like open sky.

 

Dust on the sensor would cause longitudinal lines, generally with fuzzy edges. Dust on the mirrors reduce contrast, and can cause bright areas to bloom.

 

It's theoretically possible to see Moire in fine, repetitive detail on the film. However the random nature of film grain generally acts like an AA filter with regard to Moire. Repetitive patterns would have to be comparable to sensor cell spacing, and grain size even finer, either of which is unlikely

Link to comment
Share on other sites

it is interesting that the pattern occurs only on low speed film. on high speed film with bigger grain i never seen this.

 

- I suspect that the larger grain of underexposed and overdeveloped (aka 'pushed') Tri-X will disrupt the patterning seen. Like trying to view fine detail through a nylon stocking.

 

OTOH, the more even grey expanses of the fine grain film allow those scanning artefacts to show up more clearly.

 

I'm pretty sure the effect is due to the scanner stepping becoming sticky and jerky. Thus creating repetitive small areas of overlapped or missing density.

Link to comment
Share on other sites

Moire can only occur when sampling a regularly spaced pattern, and grain is never that regular.

 

 

My best guess is scanner 'stuttering', where the scanner film transport doesn't run smoothly. Maybe it's time for a CLA?

 

 

No, it does not support ir dust removal, you are right. What is CLA?

Link to comment
Share on other sites

What is CLA?

 

- Clean, Lubricate, Adjust.

In other words, a standard service needing no new parts or extensive repair.

 

Almost any scanner that's been heavily used, or is more than a few years old, will need at least cleaning and lubrication of the transport mechanism. It has to accurately step the film by exactly one pixel's width (< 8 microns!) for every scan line. It doesn't take much dirt, dried grease or wear to disrupt that fineness of movement.

Link to comment
Share on other sites

Moire comes when the image has spatial frequencies higher than half the pixel spacing, and is

otherwise called aliasing in sampling theory.

 

There is a related effect called grain aliasing, when the grain structure has high enough spatial

frequencies, though that aren't regularly spaced. That is, the individual grains have detail

much smaller than the actual grain itself. The result is a low frequency alias of the

grain image.

 

I always scan with my scanners at the highest resolution available. Even if the scan is slow,

getting the film in, setting up to start the scan, and saving the resulting images always takes

even longer.

 

Scanning at lower resolution than a scanner is built for, that is, lower than the optical

resolution, could easily result in either Moire or grain aliasing. Color film dye clouds

don't usually have such fine structure, even when close to the resolution limit.

-- glen

Link to comment
Share on other sites

There is a related effect called grain aliasing, when the grain structure has high enough spatial

frequencies

 

- Grain aliasing never produces a repeating ripple pattern like that shown by the OP. Instead it results in a randomised exaggeration of the grain pattern, as if a far 'grainier' film had been used.

 

"Color film dye clouds

don't usually have such fine structure.."

 

On the contrary, dye clouds are micro-engineered to have a diameter of 2 to 3 microns, which is much bigger than the average halide crystal size in a comparable speed B&W film. Dye clouds also have a tendency to clump or cluster in groups of 3 to 5, making cluster diameters in the region of 9 microns across.

 

In the case of colour negative film, this can lead to an horrendously exaggerated 'grain' effect when the scanning pixel size is similar (2700 - 3200 ppi). Always provided the scanning lens also has good resolution at that spatial frequency.

 

For some reason reversal film doesn't alias to the same extent. But whatever the degree of grain aliasing, it never produces a repeating 'wavy' pattern.

 

The OP has a scanner in need of mechanical attention, and I would place good money on that.

Link to comment
Share on other sites

For negative films, the image is made up of larger grains.

 

In reversal films, it is made up from smaller grains, which didn't develop the first time.

 

Exactly what that does to the image, I don't know, but it does seem that it should be different.

 

Silver grains have structure that gives spatial frequencies much higher than the inverse

of the grain diameter.

 

As far as I know, scanners (and fixed lens digital cameras) are designed to have an

optical system with enough less resolution to avoid aliasing.

 

I mostly scan at the highest resolution that the scanner will do. I assume that

in lower resolution modes that they average pixels to reduce aliasing, but

maybe not.

-- glen

Link to comment
Share on other sites

As far as I know, scanners (and fixed lens digital cameras) are designed to have an

optical system with enough less resolution to avoid aliasing.

 

- Do you work for Epson by any chance? Because they're in the habit of quoting 'true optical resolution' (incorrectly in DPI) that the cheap lens fitted can't possibly deliver. Leading to fuzzy scans with empty resolution.

 

Decent scanners, OTOH, are fitted with lenses that match or exceed the highest PPI resolution that the sensor is capable of. Otherwise it would be pretty pointless generating that number of pixels. If grain is not imaged to start with, then it can't be aliased, can it?

 

Also, witness the number of high resolution digital cameras being marketed with no Anti-Aliasing filter fitted. The lenses used on them aren't in any way crippled to have a low resolution. So it turns out that Moire interference isn't the big boogy monster we all hid under the bed from for years, after all.

Edited by rodeo_joe|1
Link to comment
Share on other sites

Silver grains have structure that gives spatial frequencies much higher than the inverse

of the grain diameter.

Grain Aliasing is not the same as Moire. The former causes grain to appear as clumps, larger than either the grain or size of the pixels. The latter is interference between repeated patterns and the spacing of sensor cells. Grain doesn't really have a frequency, because it is random in spacing, size and shape. Under a microscope, silver appears as dendritic strands. The closer you look, the finer the structure.

 

As a corollary, the higher the sensor resolution, the smaller any Moire patterns, hence the less obtrusive with respect to the overall image. Another condition is that the lens must have greater (e.g., 2x-4x) resolving power than the sensor, which can't be taken for granted with a 40-50 MP camera.

Edited by Ed_Ingold
Link to comment
Share on other sites

- Do you work for Epson by any chance? Because they're in the habit of quoting 'true optical resolution' (incorrectly in DPI) that the cheap lens fitted can't possibly deliver. Leading to fuzzy scans with empty resolution.

From what I observe, Epson and other flatbed manufacturers quote the sensor resolution rather than the end-to-end. The figure is often asymmetric when they employ a stepping frequency greater than resolution of the sensor. This is marketing puffery, of course. Where microlenses are used, there appears to be significant overlap of the image between adjacent cells. In addition, there is a failure to consider each cell has a color filter, and the results are interpolated with a loss of spacial resolution. Interpolation alone reduces the effective resolution by a factor of about 2.

Link to comment
Share on other sites

In addition, there is a failure to consider each cell has a color filter, and the results are interpolated with a loss of spacial resolution. Interpolation alone reduces the effective resolution by a factor of about 2.

 

- No, that's a common misconception of how Bayer de-mosaicing works. Each photosite is given an equal and separate intensity weighting, which means the full true pixel-spacing resolution can be achieved. It's very easy to see this by photographing a resolution chart. With no AA filter, all bars, up to and including those at the 'pixel' spatial frequency are resolved. Digital cameras would be pretty useless for astrophotography otherwise. As it is they're the weapon of choice for amateur and professional astronomers alike.

 

"Epson and other flatbed manufacturers quote the sensor resolution rather than the end-to-end."

 

- Epson is very specific with its lies. Nearly all the published advertising bumph for their scanners definitely uses the phrase 'True optical resolution xxxx dpi'.

It's this extremely misleading and totally unsubstantiated claim that I take issue with.

Edited by rodeo_joe|1
Link to comment
Share on other sites

De-matrixing involves the analysis of a 2x2 square (Sigma excepted) to interpolate color of a given pixel. While most of the spacial resolution (luminance) is preserved, some is lost in the proess. Aliasing (errors) in this interpolation can be seen as color fringes in fine, repetitive details (fences, fabrics, etc.).

 

Epson, et. al. are not lying (except in the current, political sense), rather misleading. It is one thing to count sensor cells, and another to produce a scanned image. For what it's worth, the effective resolution of a flatbed scanner is about half that predicted by the pixel count. If you get 2200 ppi from a 4800 ppi scanner, consider yourself lucky.

 

A Nikon Coolscan has a nominal resolution of 4000 ppi, but an effective resolution of about 3500 ppi. Each line is scanned laterally, with very little overlap between pixels, then the film is advanced. Filters are not used. Instead there colored LED lamps are used, RGB + IR.

Link to comment
Share on other sites

(snip)

 

In addition, there is a failure to consider each cell has a color filter, and the results are interpolated with a loss of spacial resolution. Interpolation alone reduces the effective resolution by a factor of about 2.

 

For digital cameras, there is a Bayer array filter, such that each pixel has a filter.

 

For most scanners, that is, where either the sensor or object moves, they use a 1D array, with some choices. Some switch red, green, and blue LEDs over the array. I believe that Epson uses a Nx3 array, where the 3 is red, green, and blue. So, unlike cameras, they separately sense red, green, and blue at each position, though not at the same time. Some might scan the whole image three times.

 

There are cheaper scanners that use cellphone style camera sensors, and so do have a Bayer array on them. It is easy to tell, as it images without anything moving.

 

Also, the filter array probably only reduces by sqrt(2), the same square root when computing a standard deviation. Especially since the usual array has two green, one red, one blue.

-- glen

Link to comment
Share on other sites

From what I observe, Epson and other flatbed manufacturers quote the sensor resolution rather than the end-to-end. The figure is often asymmetric when they employ a stepping frequency greater than resolution of the sensor. This is marketing puffery, of course. Where microlenses are used, there appears to be significant overlap of the image between adjacent cells. In addition, there is a failure to consider each cell has a color filter, and the results are interpolated with a loss of spacial resolution. Interpolation alone reduces the effective resolution by a factor of about 2.

 

With some signal processing, you can extract something from overlapping scans, at the expense (there has to be one somewhere) of signal/noise.

 

For some really amazing deconvolution imaging see what they did with the Hubble space telescope before its first repair mission. For bright enough (high S/N) sources, they could get good images, as the point spread function was very accurately known. Non-linear deconvolution can extract what you might otherwise believe isn't possible.

 

But some scanners or cameras will interpolate to generate more pixels without any more resolution.

 

With electronic signals like audio, you can use a pretty sharp low-pass filter to avoid aliasing. Also with audio, you can sample at a fairly narrow window, not so far from the ideal sampling-theory delta function.

 

Camera and scanner sensors sense throughout the pixel area, far from a delta function. This is in addition to any spatial filter, commonly built from birefringent material, in front of the sensor. For fixed lens cameras and scanners, one could design an optical system with the diffraction limit appropriately selected to balance resolution and aliasing. (I don't know how close any are to doing this.) For an interchangeable lens camera, this is mostly not possible. The newer, highest resolution, cameras now exceed the usual lens diffraction limit enough that many don't have the spatial filter. (Especially at smaller apertures, which increase diffraction.) It should be fairly easy to test a scanner, scanning a resolution test pattern.

-- glen

Link to comment
Share on other sites

It should be fairly easy to test a scanner, scanning a resolution test pattern.

 

- Yep! Done that.

I have a 2" x 2" Lippmann plate carrying reproduction resolution scales ranging from 25 to 2 micron bar-space patterns. Similar to the 1951 USAF test pattern.

 

Better film-scanners that I've tested will orthogonally resolve bar-space patterns right up to their theoretical pixel spacing limit. Epson flatbed scanners won't.

 

Film scanner technology ground to a standstill some years ago when development of tri-linear CCD sensors was dropped in favour of CIS sensors for flatbed copiers. The same sensors were used in film scanners as in flatbeds, and AFAIK no film-scanner maker had/has the capability to design and manufacture their own sensors. The market just wasn't big enough to make it economical, and has now shrunk even further.

 

To the best of my knowledge, the highest resolution tri-linear sensor available has strips of 3x10,000 photosites. Thus limiting the best film-scanners to that resolution along the shortest dimension of whatever format.

 

Now 10,000 pixels per inch (or per 24mm) is obviously overkill for 35mm film, and such sensors are only used in multi-format scanners.

 

"Epson, et. al. are not lying"

 

- What else can you call a 'specification' that can't possibly be met by the equipment? Let's call a spade a spade, and a mendacious statement a lie. Not misleading, hyperbole, puff, lack of verisimilitude or exaggeration. It's just a downright lie!

Edited by rodeo_joe|1
Link to comment
Share on other sites

What else can you call a 'specification' that can't possibly be met by the equipment?

They're counting sensor cells. Are you saying that number isn't correct?

 

We've all learned (or should have learned) to hear what is NOT said as clearly as what IS said. That holds for advertising, politics, and everything else intended to entice or persuade.

Link to comment
Share on other sites

They're counting sensor cells. Are you saying that number isn't correct?

 

- I'm saying that the number of pixels captured isn't the same as "True Optical Resolution" - which is Epson's claim.

 

'True optical resolution' implies that the optical system (sensor-lens combination) is capable of whatever resolution is claimed. That just isn't the case.

 

I don't see how Epson's wording could possibly be more specific, or more misleading or just plain wrong.

 

"We've all learned (or should have learned) to hear what is NOT said as clearly as what IS said."

 

- What Epson have clearly stated is equally clearly contradicted by any review, test or serious use of their scanners. It's not a sin of omission, but just an outright lie.

 

"That holds for advertising, politics, and everything else intended to entice or persuade."

 

So the all-pervasive culture of telling mis-truths makes it perfectly acceptable then, does it?

 

I'm not naive, just totally sick of the BS crap that we're expected to unquestioningly lap up like hungry dogs!

Edited by rodeo_joe|1
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...