Mike Chaney's Tech Corner

http://www.ddisoftware.com/qimage-u

v2014.211    Apr 21, 2014

Priority: Med

v2014.211 adds raw lens corrections for the Olympus 12-40 f2.8 lens for E-M* cameras and fixes some sizing bugs WRT uncropped prints with mats.

Mike

Thanks for the lens Mike.

For the rest of you we can assure you it was in bad need of correction. It is crystal sharp but almost like a fisheye when used a 12mm uncorrected!!

Tony

In Editor (Tabs: Adjust, Levels, +Curves, Sel. Color) if I select "+Curves" check the "In" box with the up/down arrows.

I believe this box goes from blank, and 0 to 255, but the text of the numerals above 100 gets cut off making 255 look like 25 due to narrow column?


Mack

Could you show a screen snap?
Here's what I see.
Fred

Fred,

Here is the screen snap.  Windows 8.1 64-bit if it matters.

Shows 96 clearly, then maybe 125 or 128, then 140 something (144?) in the pull-down.

Odd curve in the window is something I saved to pull up the shadow detail to make a step target more linear so I use that Curve part a lot.  Otherwise it goes to black too soon.  Would be nice if there were 21 spots (Instead of Loc. 16 across as it is now) for a generic 21-step calibration tablet too.  I made a 16 step especially for Qimage, but it's not as good for setting tonal linearity as I would like.  51 steps that is also out there is too much, imho.

Mack

I get the same as Mack, with Win7 32bit. The drop down box is too narrow but the box at the top shows all three figures.
John

I get the same, now that I see where Mack is located.
Frankly, I use curves a lot, and never even tried that wheel. Never found a use for it.
I either use the eye dropper on my image to get my location and then type in a slightly higher number to lighten that location, or I just drag my points.
Those are random setting points .... Let's see what Mike says.

Fred

Fred, & others,
You ought to know nothing is random in QU  ::)
They are  the input values on the curve where a change in the setting can be made. There are 17 altogether: 0, 16, 32 ...... 224, 240, 255.
When you use the mouse pointer to change a curve, it picks up the nearest one of these points.
Thus you can set the output value for each point precisely, either by using the mouse and dragging until the desired output value is reached or by entering the number directly.
For small changes to a curve I always enter the output value directly.
Terry

Terry, somewhere down the line ask Mike about adding at least 21 steps to the LOC window over 16 please.

I noticed with the Canon 9000 II printer that only has one black ink, I have to kick the black up at the dMax (black) end of the scale pretty quick in the Qimage Curve tool or else the black shadows get bunched up in there.  The Epson that has 3-4 blacks, I can get a smoother transition into the blacks (shadows) and leave the Curve alone once made with the i1 Profiler.

If I only use the Qimage default of 16 LOC steps, I cannot get a smooth black transition with one-black ink printers even using x-rite's i1 Profiler.  A 21-step gray tablet seems to be more helpful in getting the shadows back using the Curve tool, but it is finicky to do so with only 16 settings.

You can see a bit of the step tablet in my screenshot above where it goes from black (#1) to the next LOC step 2 using the odd Curve I made up for the Canon.  Without it, the tablet is black in steps 1-3 before showing a gray.


Mack

Hi Mack,
I'm not 21 steps = 20 spaces would work as 256/20 is not whole number. 32 spaces or 33 points is the next one up I think which could make it tricky to pick with a mouse although the drop-down list could be used.
Mike does see these posts but I'll send a prompt  ;)
That sounds as though you have managed to do it? Are you entering the output values directly rather than using a mouse; that is much easier for fine adjustments?
You can then, of course, save the filter and use as a global or print filter.
Terry

Terry, yes I "generally" enter the numbers via the keyboard: LOC, and the "Out" that moves the curve to where I want it after reading it into a linearity formula in a spreadsheet.  Either way, the LOC seems to change one entire number from 0-16 (17 numbers with zero included) with the mouse as well.

I tried to do something like 4.5 for a LOC which Qimage accepts, but it seems to round down to 4 doing that on the curve's line itself and wipes out my "Out" setting I used for LOC 4.  Even 4.9 rounds down to 4 on the curve too.  So I guess 0-16 is it and nothing else?

Most of the old Kodak and Stouffer calibration gray scale step-wedge targets use 21 steps of 5% each to set up print ink linearity.  The Qimage 16 step LOC seems odd, excluding zero is added making it 17 steps.  I'm not getting the tones to separate it that well in the dark areas with only 16 numbers.  Shadows, with the single black ink printers, seem to lose detail without fiddling with the 0-3 area of the LOC for me in Curve part of Qimage.

You can see in the screen capture above where the curve I used lifted the #2 out of darkness with the steep curve off the zero point (Made for the Canon printer.  Epson is fine, but it has more black inks too.).  Another LOC point down there would help.

Fwiw, "How to make and print a 21 step tablet in PS" is here: http://www.jnevins.com/stepwedge.htm (http://www.jnevins.com/stepwedge.htm)  My Canon is showing me some odd colors in the middle too (yellow in some steps and magenta in others), but I could use the Curve and negate them in the RGB part individually I guess.  Epson is fine and seems to point to the need for more black inks too.

Mack

LOC is "Location number", ie. the position along the horizontal axis so they are whole numbers only and correspond to the fixed input values 0, 16, 32 ...... 224, 240, 255.
Terry

 ???

In the LOC box (horizontal), I cannot enter 255, only goes up to 16?  Rounds down too for any whole number placed there too (e.g. 4.9 rounds down to 4, etc.).

The other boxes (Especially the OUT) I can which move the vertical part of the curve form 0-255 (Which makes that 0-16 LOC number on the curve darker or lighter in the print.).

***********************

Aside, I see today Keith Cooper released a new B&W Test Image with 51 steps of gray http://www.northlight-images.co.uk/article_pages/bw_printing/bw-test-image-2.html (http://www.northlight-images.co.uk/article_pages/bw_printing/bw-test-image-2.html), or another with 21 steps to set up linearization curves.  That's why I would like to see a LOC ability to set 21 steps instead of 16 since 21 seems to be some industry standard set by Kodak and Stouffer in the past.  51 is a bit much to put into Qimages Curves part manually - beside it not being able to do more than 16 in current state.

I also noted from the new test image above when I tried it today, the Epson printed and centered it fine in Qimage.  The Canon 9000 II shifted downwards a bit and the border was off at the end.  However, I think the Epson 3880 has a better sensor for the beginning of the paper feed so it might be a Canon engineering goof.  The Canon also didn't do well on the faded black spot near the A4 in the new K.C. test image either and produced some bad banding there where the Epson was smoother.  The Canon might go into the trashcan soon I I just replaced the head too and it still does it.  I suspect the Canon Pro-100 with the other two gray/black inks does better on B&W than the 9000 II which had a Green and Red ink and seems the Pro-100 is the same model, just they ditched the green and red for the two grays to make the gray step transition smoother (or do the "Curve kick" like I had to do above in Qimage.).


Mack

LOC stands for location - its a number, that is all and you are not meant to enter a number there. The Location along the horizontal axis is selected from the In drop-down, each position along the axis has a number 0 to 16, it's not rocket science  ::)
Terry

Terry, I got that 0-16 LOC X-axis stuff.  However, it's too small of a range.

What I'd like to propose is at least 0-21 for that box so I can use the generic industry standard 21-step gray scales (Kodak & Stouffer) and now Keith's 21-step wedge to set the values read from those in Qimage.  I think even my i1Profiler uses a 21 step tablet in its Colorport section to read from.  With Qimage locked into such a small LOC value range of 0-16, I cannot do so.  As it is, blacks get bunched up and it's too hard to separate them with such a small 0-16 range to work within, so it needs a finer (or more available LOC numbers) range to work with.

I tried to do so by fooling it into a LOC=4.5 but it rounds down to 4 so I am stuck with 0-16 for any and all X-axis movements.  I need to apply more than 16 sets to the curve from these wedges to smooth out the tonality and make it linear.

I don't know what to make of the 51 step gray wedges, but they seem to be showing up more in digital land.  Especially in setting up the the 8 tanks of differing black and gray piezo inks.

Mack

We have 17.  21 only gives you 4 extra steps and those will be spread out over the entire range.  So if we're talking blacks, that's the bottom 25% of the range (at most).  So you might get one extra step in the range you are working if it was 21 instead of the existing 17.  You really can't make that work?  It is, after all, a smoothed curve.  Meaning that you can move the point just before or just after the missing single entry you have in the 21 step curve and the "missing" step will move accordingly.  If you have a curve so jagged down in that range that you need a curve that looks that erratic, you probably need to solve the underlying problem that is causing the jagged tone curve at the source.

Mike

Thanks Mike.  We're on the same page at least on the LOC numbers (0-16).

If there was an easy way to call in the 21 steps to the LOC to make it linear it would be great.  As it is, the spreadsheets that do this L*ab linearization stuff automatically take some step, like step 19 that goes off the straight line where all the others are dead on.  Then I have to try and figure out that gray scale step 19 (if it is towards the white end) a LOC 14 or 15 in Qimage's LOC's box?  It's really 14.48 if my math is right, but I cannot plug that in to the LOC as the scale is too small and fixed on whole numbers both.

If I could set say LOC=14.5 that would work, although takes me some math to get from that step tablet number of 19 into Qimage.  As it is, Qimage rounds down to 14 and it messes up that number which was correct based off the spectrometer's readings so that part of the curve now gets messed up when it was correct.  For piezo inks, it becomes a mess since it pulls in another gray ink tank out of 8-9 tanks.

Actually, even on a 51 step scale, if the linearization spreadsheet says step 49 is off, if I could plug in 15.4 into Qimage that would work even with the 0-16 current range in it now.  Problem with the 51 steps if a range of say 44-49 is off, then how to null it out without using a finer scale in the LOC axis range?

Some of these black inks are very finicky in the Qimage Curves I find as they also call in other colors to make them neutral.  That fading blot of Keith Cooper's newest B&W Test Image sure shows up banding if the tuning curve isn't spot on.

Aside, QuadToneRip (QTR) doesn't work well in newest Windows and Roy seems to have abandoned it to Apple/Mac as his Windows coder never updated his compilers for newer Window versions.  It allows for a lot of curve tuning in blacks and colors of ink, I think it did more than 51 steps and maybe more as it pulled in the spectrometer readings and adjusted the curve automatically.  Just it crashes a lot in Windows 8 and Roy cannot fix it so it is a very old version now.  Windows needs a new QTR B&W printer software of some sort (hint. hint.  ;) ).

Anyway, food for thought.  0-16 with the ability to set 1.3 (example) would be better for fine tuning with these new 21 & 51 step tablets, just a bit more math work or a spreadsheet formula to correct for entry into Qimage LOC box.

Mack

Well, assuming the 21 step scale goes from 0-20, here's some math that will get you pretty close:

P (uppercase) = 17 step
p (lowercase) = 21 step

As you can see, 0, 4, 8, 12, and 16 on the 17 step scale align perfectly with values from the 21 step scale.

P0: p0
P1: p1*0.75 + p2*0.25
P2: p2*0.50 + p3*0.50
P3: p3*0.25 + p4*0.75
P4: p5
P5: p6*0.75 + p7*0.25
P6: p7*0.50 + p8*0.50
P7: p8*0.25 + p8*0.75
P8: p10
P9: p11*0.75 + p12*0.25
P10: p12*0.5 + p13*0.5
P11: p13*0.25 + p14*0.75
P12: p15
P13: p16*0.75 + p17*0.25
P14: p17*0.50 + p18*0.50
p15: p18*0.25 + p19*0.75
p16: p20

Example: to get loc 3 on the 17 step scale, take loc 3 on the 21 step scale times 0.25 and add loc 4 on the 21 step scale times 0.75.

Hmm.  Now that I think about it, lemme know if this works.  If so, I could probably build in a 21 step converter feature where you can enter all 21 steps and it'll automatically make the QU 17 point curve for that.

Mike

Mike, if I read your math right, the small "p" is off my 21 step chart.  Those are fixed whole numbers so Step 1 is p1, Step 2 is p2 etc. on the X-axis.

If I do the table I get this:

P0: p0                 = 0
P1: p1*0.75 + p2*0.25   = (0.75)+(0.50)  P1=1.25   i.e. (Step 1*0.75) + (Step 2*0.25) for a Qimage LOC # = 1.25   ???  Can't input that into QU.  Rounds down to 1 only.
P2: p2*0.50 + p3*0.50   = (1.0)+(1.50)   P2=2.50    Drops down to LOC=2
P3: p3*0.25 + p4*0.75   = (0.75)+(3.00)  P3=3.75   Drops down to LOC=3
P4: p5         =    = P4 or LOC4 = Step 5
P5: p6*0.75 + p7*0.25   = (4.5)+(1.75)    P5=6.25

... etc.

Seems to be dependent on some other step than the single one I want to address, and also altering its neighbor, aside from the decimal part?

Somehow I can't see how 21 'whole numbers' off a step wedge reading can be fitted to 17 'whole number' LOC steps without some sort of decimal in the mix too?

Was trying to do this in a spreadsheet and got a math headache.  Ugh!  :D

______________________

Aside:

I'll attach a sample 21 step wedge (converted) JPG, but I need to reverse the Black to Zero and White to 21 for use in Qmage.  What usually happens with some cheap printers the blacks get bunched up around the 18-21 scale in the JPG.  I think my Canon 9000 II falls into the "cheap" category (Hence the need to do a quick bump off the left side of the Curve in QU as above), and the Epson 3880 addresses the issue a bit better with more balck shades of in.  Tuning the black shadows so they separate in the cheap Canon is a problem, and the i1 spectrometer tells me its blocked up too.  Just cannot refine it as well since the QU LOC range/scale is too compressed.

Don't know why 51 gray scale wedges is some big deal either, but those are out there.  Might be the 8-9 blacks in piezo inks may make the same "bunching up of black shadows" issue with a smaller 3 black inks 'color' printer; Somewhat like Canon 9000 II is to 3-black ink 3880 (Excluding the other Epson 3880 Photo Black or Matte Black since only one can be used at a time, along with the other Light Black and Medium Black.).

Keith Cooper's newest B&W Test Image has the 51 step wedge that is capable of being scanned with the i1 ColorPort software which might make it easier than reading all 51 steps individually.  Should be easier to dump the readings into a CRV file to read in the software or a spreadsheet that shows the linearization curve.  He also did a 21 step as well maybe for manual entry or the cheaper ColorMunki Photo.  Must be some standard since Kodak had 21 step wedge for decades, and Stouffer in the printer biz both.

Mack

No.  Capital P is the value on your 17 step curve.  Lower case p is the value on your 21 step curve.  So figure up your 21 step curve and convert to the 17 step curve using the formulas.

If your 21 step curve starts with these values:

p0 = 0
p1 = 12
p2 = 26
p3 = 38
...
p20 = 255

Then the 17 step curve would look like:

P0 = 0     [p0]
P1 = 16     [12*0.75 + 26*0.25]
P2 = 45     [26*0.5 + 36*0.5]
...
P16 = 255

Mike