denoir wrote:
Probably just different color profile for the 5DII RAW development. The first shot is just resized and all sharpening operations are done on the L channel alone which means that colors can't be affected.
Could you make a version available of the image after RAW development and before the downsampling process?
The cause of the green becomes pretty obvious when you look at it 100% - it's simply LoCA green fringing.
There is a little green fringing in the original, but it seems to be enhanced by your processing - or reduced by mine. The resulting blinds end up less green than yours:
Well yes, mine is based on the premise that you want to amplify the fine detail so that it gets preserved when you resize the image. Yours was a low pass filter - i.e it removes the fine detail in the image. So the results are not that surprising.
Toothwalker wrote:
Your first and second version have in common that the blinds end up greenish, whereas I get gray blinds. I suppose that is a bug/feature of the change to Lab color mode?
Using RGB mode should worsen the green color because sharpening will also boost color contrast (saturation).
One thing I'd change in your version is mask the inside of that circle between 5 and 6 o' clock because there's a very obvious stairstep due to aliasing.
denoir wrote:
Well yes, mine is based on the premise that you want to amplify the fine detail so that it gets preserved when you resize the image. Yours was a low pass filter - i.e it removes the fine detail in the image. So the results are not that surprising.
Well, my last version was just a bicubic resize followed by a high-pass sharpen. I am getting interested in the subject. I wonder, does your image really have more detail? (By the way, your second version is more aggressively sharpened than your first one, so I don't think that you applied the same resize method after blurring the blinds in the circle.)
Toothwalker wrote:
I wonder, does your image really have more detail?
Well, clearly, since as you observed it shows the green fringing while it is lost in your version. That was why you started this particular discussion, wasn't it?
(By the way, your second version is more aggressively sharpened than your first one, so I don't think that you applied the same resize method after blurring the blinds in the circle.)
My second version has been modified since I first posted it. So yes, it's not the same script.
denoir wrote:
Well, clearly, since as you observed it shows the green fringing while it is lost in your version. That was why you started this particular discussion, wasn't it?
Yes, but in the process I started to wonder about more things. Your versions preserve the green of the fringing, not the fringing itself since that corresponds to a spatial frequency that cannot be represented in the smaller images.
I guess there is something wrong with my bicubic interpolation. Other methods do deliver the green, although your RAW development also helps. There are several factors.
My second version has been modified since I first posted it. So yes, it's not the same script.
And now your first version has also been modified? (The sharpening is a bit too aggressive for my taste.)
The first version was modified at the same time as the second version. But never mind that - it was never intended to be a general sharpening comparison but just to deal with the moiré issue. Forget about that example. I've got something better.
I generated an image in Matlab with four sets of line pairs - 10 lp/mm, 20 lp/mm, 40 lp/mm and (sensor max) lp/mm. I used the M9 sensor as a model - i.e 24x36mm that produces 5212x3468 pixel images. Sensor max in that case is 72 lp/mm. For each set the lines go from 50% gray to black.
denoir wrote:
The first version was modified at the same time as the second version. But never mind that - it was never intended to be a general sharpening comparison but just to deal with the moiré issue. Forget about that example. I've got something better.
I generated an image in Matlab with four sets of line pairs - 10 lp/mm, 20 lp/mm, 40 lp/mm and (sensor max) lp/mm. I used the M9 sensor as a model - i.e 24x36mm that produces 5212x3468 pixel images. Sensor max in that case is 72 lp/mm. For each set the lines go from 50% gray to black. Here's the full image - run your script on it. ...Show more →
Very good.
Here's my script vs bicubic resize (open in separate tabs in the browser and flip between them):
Okay, let's get more serious about setting a date for the next meeting. Things are getting more complicated though. Here are the next few weekends so far:
September 10: Jonas B, carstenw, -Morfeus
September 17: carstenw, -Jonas B
September 24: carstenw, -Jonas B
October 1: carstenw, -Jonas B
October 8: carstenw, Jonas B
Jonas, I presume you meant October 8, not November 8, which is a Tuesday?
November sounds a bit cold... Checking my autumn schedule I'm sure I meant to say Saturday October 8.
October 8 sounds a bit more realistic than September 10, it's easier to get a cheap flight if booking a little in advance. If we decide quickly September 10 can work as well. But missing Heinz? Do you usually have sparkling autumn leaves in in October, perhaps in Tiergarten and other parks?
September 10: Jonas B, carstenw, -Morfeus, -Ulff
September 17: carstenw, -Jonas B, -Ulff
September 24: carstenw, -Jonas B
October 1: carstenw, -Jonas B, -Ulff
October 8: carstenw, Jonas B, denoir, philber, -Ulff
October 15: carstenw, Ulff
We definitely have problem finding a date where everyone can make it. If both Jonas, Luka, Philippe and Heinz can make it on October 8, then I suggest that we choose this as one of the dates. There is nothing stopping us from adding other dates though, either earlier or later or both. It is helpful to decide in advance though for those who want to try coming from other countries.
Jonas, the leaves are okay here, nothing like as spectacular as Canada (where I lived before Germany). There are many interesting places to go though, and the foliage should still be relatively intact, I believe. We also have nice indoor venues, if anyone wants to try that. The Technikmuseum Berlin is great, for example, with steam locomotives and suspended classic airplanes. Lots of interesting detail work. Tiergarten is probably still beautiful, perhaps also Botanischer Garten, and the architecture can of course be done any time of year. We could try to calculate when the blue hour (dusk) is, and line up a nice evening shot for that.
What would people like to do?
Markus, is there any chance you could have October 8 free?
People, there is simply no way that this sort of meeting can take place if there is a pre-condition that everyone must be there every time. Besides, as we are weeks away from any date, things can well happen that can change the list of attendees.
As matters stand it seems that two dates are promising. I would suggets going ahead with both dates. That way, nobody gets left out, and those whould could, at this stage, make both (Jonas, Carsten, Ullf and I) get to decide which one they will attend, or both. And any meeting that has a list of 5 members minimum (or whatever number you think is good, but 5 looks OK to me) goes ahead...
So, what do you think?
More people might read it too.