Axial Lights

[mjkzz]

Here is a comparison between co-axial lighting and dome + tube lighting. Co-axial reveals much more scratches and sensitive to curvature on the surface, lights fall off faster at edges. The flat profile dome light gives is probably good reading stuff off a shiny surface. For computer vision applications, I like the co-axial one as it catches defects on parts -- look at where red arrows pointing to, the one lit by dome light is hard to distinguish.

dome lit

co-axial lit

[mjkzz]

Dome (or hybrid dome because I have a tube around the coin) light definitely produce much flatter image, compare this one with the one from previous post, look at the letter D, O, S or any of the letters, they appear much flatter, almost as if they were machined so. On the other hand, the co-axial lights captures the curvature much better, and those lettering appear rounded.

For those curious, here is the TIF file for this image, check it out. https://www.dropbox.com/s/v855n5powc5dw ... d.TIF?dl=0

OK, I am out, back to work

[mjkzz]

oops, forgot to post it here for the HK 10 dollars coin. 6 stacked images with 18 images in each stack, about 2x (probably 1.7x) with Pro Optics 12K line scan lens (I think I have posted info about it here, really sharp one, looking for a 16k similar lens). Since it only has 18 images for each stack, easy to handle, so I just manually moved the coin for the 6 frames, not that painful at all. The key part here is the LONG working distance of the lens, well over [EDIT] 120 mm even at 1.7x magnification.

HK 10 Dollar Coin On EasyZoom Check out all the details.

Line scan lens used:

[jurkovicovic]

I would like to ask you.
I guess the good results didn't come right away. Did you have problems with image sharpness during the tests?
I'm quite interested in this topic. I tried a few photos, but I'm having problems.
One part of the image is sharper than the second one. Do you have any information on this matter?

[mjkzz]

edited post -- working distance is about 120 mm(not 200mm) at 1.7x for the lens, just not to mislead anyone A 2x 16K no name line scan lens is coming, front element is chipped but seller is so nice to let me try it (knew him for 5 years now).

[ray_parkhurst]

Image on EasyZoom looks good.

I did a search for the 12k scan lens but no info I could find (though I am not the best searcher...). Do you have info on that lens?

[ray_parkhurst]

I would like to ask you.
I guess the good results didn't come right away. Did you have problems with image sharpness during the tests?
I'm quite interested in this topic. I tried a few photos, but I'm having problems.
One part of the image is sharper than the second one. Do you have any information on this matter?

Which image are you referring to? Your image or one of mjkzz's?

[mjkzz]

Image on EasyZoom looks good.

I did a search for the 12k scan lens but no info I could find (though I am not the best searcher...). Do you have info on that lens?

I think it was back in late 2018. I think it is made by a Korean company, Pro Optics. It is really sharp and since it is relatively unknown, it was cheap. At the time, I had the choice of Linos 105 which costed about 10 times more, so I gave it a try and it worked. Also, I do not know working distance of the Linos 105 and that is a factor -- I was using a spherical diffuser and I need LONG working distance.

Here is an 8081x8081 image of a brand new coin.

part of that image at 100% crop.

[mjkzz]

Replying to this, I just realized I do not have a full frame camera (currently, I have Sony A7III, 5.9um pixel pitch) that can match the incoming 16K line scan lens, so I can not take full advantage of it. Nuts Excuse to get a Sony A7RIV

[ray_parkhurst]

Replying to this, I just realized I do not have a full frame camera (currently, I have Sony A7III, 5.9um pixel pitch) that can match the incoming 16K line scan lens, so I can not take full advantage of it. Nuts Excuse to get a Sony A7RIV

Well, if you end up getting an A7Rm4, I'll leave it to you to figure out a practical stack and stitch workflow. So far the only thing I've been able to do is single 4-shot images, and very limited stacks due to the onerous nature of the workflow. Luckily the A7Rm4 has a high quality jpg engine that outputs a very nice jpg in "fine" mode. I've been doing 2-tile stack and stitch in jpg mode and it has exceeded my expectations, including much less false color than expected without pixel-shifting. But it would indeed be nice to have a practical workflow and since I'm using your stack and stitch controller, I'm hoping you can figure one out

BTW, what are the max apertures for the 12k and 16k lenses?

Edited to add: I'm seeing some unnatural looking shapes in the above 100% image. They are similar to what I saw with my early pixel shifting experiments with the Pentax K3 and the Sony A7Rm3 when using Imaging Edge. I attributed it to oversharpening not at finest detail level but at intermediate feature sizes. Do you see what I'm talking about?

Further edit: I see that there are areas of the coin that have an intentional texture. Are these letters intentionally textured?

[mjkzz]

I had a blog on that lens, Here it is. But for some reason, I could not get to their website for specs.

In the future, since coin collection was my childhood hobby (sort of), I might automate it completely, user puts the coin in a slot, then machine does the rest (finding outline, stacking range, etc), sort of like "coin photography for dummies"

But in near term, I find it painless to do, even with 6 frames/tiles at 1.7x. However, I just got a co-axial device for objectives, with that, say at 5x, number of images increases exponentially (almost cubed), that will be painful and need automated process, but still manageable with SnS system.

Yes, I do notice some stuff in the final image, I think it is due to co-axial lighting, how the coin was made (lots of patterns on the surface of coin), and yes, sharpening. Here is 100% crop of one image in one of the stack, it shows line patterns that might seem un-natural - under co-axial light, even slight depth difference (scratches, etc) will be "amplified" and shown. Then these "scratches" get sharpened and they become even more 'un-natural' because co-axial light is not natural to us.

That is my interpretation.

[jurkovicovic]

I would like to ask you.
I guess the good results didn't come right away. Did you have problems with image sharpness during the tests?
I'm quite interested in this topic. I tried a few photos, but I'm having problems.
One part of the image is sharper than the second one. Do you have any information on this matter?

Which image are you referring to? Your image or one of mjkzz's?

Ray I´m writing about my output. Here is a demonstration.
The star on the right is relatively sharp, but the star on the left is not sharp.
I have no idea where I should look for a problem.

100% crop, edited.

[lothman]

Replying to this, I just realized I do not have a full frame camera (currently, I have Sony A7III, 5.9um pixel pitch) that can match the incoming 16K line scan lens, so I can not take full advantage of it. Nuts Excuse to get a Sony A7RIV

That is the Laowa 100mm Macro lens at f4, stacked withe the Sony A7riv, Zeren->Topaz denoise (clear standard) -> Photoshop (contrast, curves), magnification around 1:1. I don't think a linescan lens will improve, and the Laowa gives you up to 2:1.

and the crop (heavy jpg-compression to match forum limit)

same area but at 1,7x magnification same lens settings

[mjkzz]

thanks lothman, I know Laowa 2x is sharp, I had one but sold it because I know it is a very good one, but my line scan lens was about 100USD or maybe less, too long ago.

[mjkzz]

Further edit: I see that there are areas of the coin that have an intentional texture. Are these letters intentionally textured?

I think so, particularly the numbers (1 and 0), there are also some counterfeit proof markings inside those numbers -- if you view it from top to bottom, it reads 10, from bottom to top, it reads RMB (abrev for Chinese currency)

If you check out the HK 10 dollars coin image on Easyzoom, you do not see "weird" shapes nor artifacts even though it was sharpened the same amount (I think it is over sharpening, learning how to use Lightroom). One fact is that the HK 10 dollars coin was lit differently from co-axial (though it looks similar to coaxial). Another guess is that the coin has microscopic patterns that show up when lit coaxially and sharpened.