Polarized lighting setup, recommendation to improve?

[Michael G.]

Hi,

I'm trying to improve my lighting setup at the moment.
Using polarizing filters seemed to be promising, but I don't get them to work to even light fiducials on several colors of pcbs and to detect the holes of tape stripes.

please see attached images to get an overview about my setup. in the other image I put some snapshots taken on a white silkscreen pcb, green silkscreen pcb and at grey feeder. the glare at the feeders is removed perfectly, but on the pcbs the fiducials are not good to be ssen.

I would be interested if anybody had put polarizing filters into service and how? Or am I on the wrong way here? What have you on your machines and how perfect works it for you? I think I would need some coaxial light?!

[Cri S]

You should invest into light foils or Vreg for dimming.
Polarizer remove only reflections, you have bigger problems.
Beside reflection, the Pins from K4 component need to be clearly
delimited without
polarizer. If you don't resolve that, no filter can help you universally.
On the green pcb, the vac sensor need to have the pins illuminated on
both sides, not
only on the right side. Probably it's a optical axis problem, but you
should place alu foil
and adjust the angles of iluminations in order that you have equal illumination.

2017-09-18 20:06 GMT+02:00, Michael G. <mic...@groene.de>:
> Hi,
>
> I'm trying to improve my lighting setup at the moment.
> Using polarizing filters seemed to be promising, but I don't get them to

> work to even light fiducials on several colors of pcbs *and* to detect the

> holes of tape stripes.
>
> please see attached images to get an overview about my setup. in the other
> image I put some snapshots taken on a white silkscreen pcb, green
> silkscreen pcb and at grey feeder. the glare at the feeders is removed
> perfectly, but on the pcbs the fiducials are not good to be ssen.
>
> I would be interested if anybody had put polarizing filters into service
> and how? Or am I on the wrong way here? What have you on your machines and
> how perfect works it for you? I think I would need some coaxial light?!
>
>

> --
> You received this message because you are subscribed to the Google Groups
> "OpenPnP" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to openpnp+u...@googlegroups.com.
> To post to this group, send email to ope...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/openpnp/f07bf32b-1419-4b97-b65b-efced071a0a6%40googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
>

[Cri S]

If you are not able to resolve it, mask it out, rotate the strips 90
degree and work with that.
Just use mask inside pipeline because the bigger area even if it's
useless for vision is
nice to have for setup parts.

[Michael G.]

hi cri s,

was hoping for some advice, thank you. have some further questions, written into text below.

Am Montag, 18. September 2017 21:15:45 UTC+2 schrieb Cri S:

You should invest into light foils or Vreg for dimming.

what are light foils for? you mean diffusing foils or just some less permeable to dim manually? or do they condition the light in another way?

have different led stripes with different power level here. but dimming would be good i think? any ic-device you recommend not to get into issues with flicker?

 

Polarizer remove only reflections, you have bigger problems.
Beside reflection, the Pins from K4 component need to be clearly
delimited without
polarizer. If you don't resolve that, no filter can help you universally.

have you an idea on how to resolve that? have read in a handbook (isbn 978-3-00-051330-5) about many different lightings. dome, flat dome, bright light, dark light, diffuse, coaxial lighting. I am a little confused what to use. just direct light brings so much reflections.

 

On the green pcb, the vac sensor need to have the pins illuminated on
both sides, not
only on the right side. Probably it's a optical axis problem, but you
should place alu foil
and adjust the angles of iluminations in order that you have equal illumination.

attached a picture of my lighting. how to i find the correct angle? is plain better? or a bit steeper?

where shall I place alu foil? lay out the table with it to check the illumination and reflections?

 

thank you

[Michael G.]

just placed alufoil on the table, see attached. the shadow in the middle is going with the cam if moving. tried to get it away with another led stripe I hold in different angles next to the camera but the shadow remains. it's an elp camera with 6mm lens. maybe its the lens?

[Michael G.]

it seems not to be the lens, tried another one. that would have been too easy ;)

[Marek T.]

As understood you have the stripes from two camera's sides. Probably better effect you could get if arround the lens with diodes to light up the center (where the camera's focusing) from every sides. Changing the height position of diodes (distance from them to pcb) you will have more or less visible shadow effect, I think you have the shadow made by the lens body.

[Cri S]

2017-09-18 21:33 GMT+02:00, Michael G. <mic...@groene.de>:
> hi cri s,
>
> was hoping for some advice, thank you. have some further questions, written
>
> into text below.
>
>
> Am Montag, 18. September 2017 21:15:45 UTC+2 schrieb Cri S:
>>
>> You should invest into light foils or Vreg for dimming.
>>
>
> what are light foils for? you mean diffusing foils or just some less
> permeable to dim manually? or do they condition the light in another way?

The standard one diffuses the lights making it uniform. It uses
lithographic tecnology
for the master. Nanooptics.
The minimal cheapest alternative is white baking paper and/or paper
with low plaster, two sheet between frosted (sanded with sand paper)
plexiglass if you not have cheap source of frosted plexiglass. As
example the cd doom
makes good diffuser material..

Just a example that let you give some hints about light foils, not as
i would suggest
this product in particular or have used any of this brand.

> have different led stripes with different power level here. but dimming
> would be good i think? any ic-device you recommend not to get into issues
> with flicker?

I recommend lm317 , but resistor and voltage should be matched for
greater control.
Otherwise higher speed pwm works using uC, ,

>
>
>> Polarizer remove only reflections, you have bigger problems.
>> Beside reflection, the Pins from K4 component need to be clearly
>> delimited without
>> polarizer. If you don't resolve that, no filter can help you universally.
>>
>
> have you an idea on how to resolve that? have read in a handbook (isbn
> 978-3-00-051330-5) about many different lightings. dome, flat dome, bright
> light, dark light, diffuse, coaxial lighting. I am a little confused what
> to use. just direct light brings so much reflections.
>

It depends what you want to do. If you have the need to read the
marking signs too
in order to check correct part or not.
There are a lot of possibility, and normally none works without
dimming to my knowledge.

>
>> On the green pcb, the vac sensor need to have the pins illuminated on
>> both sides, not
>> only on the right side. Probably it's a optical axis problem, but you
>> should place alu foil
>> and adjust the angles of iluminations in order that you have equal
>> illumination.
>>
>
> attached a picture of my lighting. how to i find the correct angle? is
> plain better? or a bit steeper?

It depends. Place the light near the camera, or far away.
First build dimmer, then you need to timing the camera with light for
the required
25/30/... frames/seconds and then you could proceed with the leds you
have to see
what works better.

> where shall I place alu foil? lay out the table with it to check the
> illumination and reflections?
>

take care removing the foil from roll and use tape to wrap it around
pcb and tension it.
It is then like a mirror and you could check it visually. Later you
can add baking foil
and check better the result using thresholding.

> thank you
>
>>
>> 2017-09-18 20:06 GMT+02:00, Michael G. <mic...@groene.de <javascript:>>:

>> > Hi,
>> >
>> > I'm trying to improve my lighting setup at the moment.
>> > Using polarizing filters seemed to be promising, but I don't get them to
>> >
>> > work to even light fiducials on several colors of pcbs *and* to detect
>> the
>> > holes of tape stripes.
>> >
>> > please see attached images to get an overview about my setup. in the
>> other
>> > image I put some snapshots taken on a white silkscreen pcb, green
>> > silkscreen pcb and at grey feeder. the glare at the feeders is removed
>> > perfectly, but on the pcbs the fiducials are not good to be ssen.
>> >
>> > I would be interested if anybody had put polarizing filters into service
>> >
>> > and how? Or am I on the wrong way here? What have you on your machines
>> and
>> > how perfect works it for you? I think I would need some coaxial light?!
>> >
>> >
>> >
>> > --
>> > You received this message because you are subscribed to the Google
>> Groups
>> > "OpenPnP" group.
>> > To unsubscribe from this group and stop receiving emails from it, send
>> an

>> > email to openpnp+u...@googlegroups.com <javascript:>.

>> > To post to this group, send email to ope...@googlegroups.com

>> <javascript:>.

>> > To view this discussion on the web visit
>> >
>> https://groups.google.com/d/msgid/openpnp/f07bf32b-1419-4b97-b65b-efced071a0a6%40googlegroups.com.
>>
>>
>> > For more options, visit https://groups.google.com/d/optout.
>> >
>>
>
> --
> You received this message because you are subscribed to the Google Groups
> "OpenPnP" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to openpnp+u...@googlegroups.com.
> To post to this group, send email to ope...@googlegroups.com.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/openpnp/c825c673-a259-4999-a428-5a82d14adfd3%40googlegroups.com.

[Cri S]

The alu foil need to be undamaged and tensioned.
It is a cheap substitute of surface mirror.
If it's a problem for you, use blank paper and doing
histogram-equalize after converting it to
grayscale, or use threshold on gimp as example using the slider to see
the results.

[Michael G.]

thanks for all your hints, will reread tomorrow and try to find the best result for me.
in the meantime I printed a tunnel dome. for me as novice in computer vision the results seem much better. the only thing is a hole in the middle, but this is a limit of the principle (coaxial lighting in same intense would avaid that).
what do you think? I'm going to make the hole for the cam smaller to the minimum required and try to optimize that design and build the dimmer...

[Cri S]

The hole is ok, if you make it significant smaller you can have
problems with fiducials.
Dimming the leds make it better, for tests the two stripes should be
dimmable separatly.
Further take a photo of white paper and check threshold, and attach it
as picture without
cropping.
The biggest problem remain, you have a well defined area where
illumination works great,
and outside it fails. It seems that the area is 24x24mm . Is this
enought for you ?
It is more a issue for loose part feeder, dimming could help and it is
worth checking ligthning
on blank paper.

2017-09-18 23:13 GMT+02:00, Michael G. <mic...@groene.de>:
> thanks for all your hints, will reread tomorrow and try to find the best
> result for me.
> in the meantime I printed a tunnel dome. for me as novice in computer
> vision the results seem much better. the only thing is a hole in the
> middle, but this is a limit of the principle (coaxial lighting in same
> intense would avaid that).
> what do you think? I'm going to make the hole for the cam smaller to the
> minimum required and try to optimize that design and build the dimmer...
>
> Am Montag, 18. September 2017 22:50:20 UTC+2 schrieb Cri S:
>>
>> The alu foil need to be undamaged and tensioned.
>> It is a cheap substitute of surface mirror.
>> If it's a problem for you, use blank paper and doing
>> histogram-equalize after converting it to
>> grayscale, or use threshold on gimp as example using the slider to see
>> the results.
>>

>> 2017-09-18 22:43 GMT+02:00, Cri S <phon...@gmail.com <javascript:>>:
>> > 2017-09-18 21:33 GMT+02:00, Michael G. <mic...@groene.de <javascript:>>:

>> https://groups.google.com/d/msgid/openpnp/c825c673-a259-4999-a428-5a82d14adfd3%40googlegroups.com.
>>
>>
>> >> For more options, visit https://groups.google.com/d/optout.
>> >>
>> >
>>
>
> --
> You received this message because you are subscribed to the Google Groups
> "OpenPnP" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to openpnp+u...@googlegroups.com.
> To post to this group, send email to ope...@googlegroups.com.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/openpnp/eeeb7b0a-e293-481c-b4fc-41aab2f7eeaf%40googlegroups.com.

[Michael G.]

Hi,

tried my next version of the tunnel dome (see attached stl). improved the length to have a wider lighted area as well as I made the hole for the camera smaller.
I think I can live with that now. the only problem would be on very glossy pcb pads: see fiducial_(off)center(ed) and compare. the camera is mirrored in the pad/fiducial. I think that can not be resolved without much effort/buying coaxial lighting or a flat dome so I would take as is. furthermore the fiducial on the pcb has not enough free area without soldermask which is suboptimal for vision.

the image with the histogram equalized shows vignetting to the edges, but that I will just crop if it causes trouble and it's looks not too bad for me.

ordered dimming modules (http://www.ebay.de/itm/5x-DC-Spannung-Step-Down-Schaltregler-Netzteil-Modul-Wandler-Konverter-LM2596S/182452589262?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2057872.m2749.l2649) and will see, if I can improve lighting further.

[Michael G.]

some more images from the feeders.

[Cri S]

It seems that the camera calibration should be improved.
Canny circle detection works without problems.
If dimming and changing gamma (contrast) it should work.
For using it on loose feeder you should place 4 direct downlooking
leds (normal leds)
at the 4 edges in order to reduce vignetting, it is not optical, it's
induced from lightning.
Eventually you should check if placing a baking paper at 30 degree
over the leds helps
for the center spot illumination drop. This after dimming module are installed.
Example of fiducial preparation, with steps for visual checking, it
works good for
hough circle, don't know for template match, i see no reason that it
don't work if adding
masks, hovewer the holes near the fiducial can have better match on
templatematch.
As you could note , with less illumination it works fine, the other
fiducial have too mutch
illumination and it is detected wrong (mask detected, not hole). And
if you look on orginal
image the metal is better as the fiducial near R3.

2017-09-19 21:36 GMT+02:00, Michael G. <mic...@groene.de>:
> some more images from the feeders.
>
> Am Montag, 18. September 2017 20:06:30 UTC+2 schrieb Michael G.:
>>
>> Hi,
>>
>> I'm trying to improve my lighting setup at the moment.
>> Using polarizing filters seemed to be promising, but I don't get them to

>> work to even light fiducials on several colors of pcbs *and* to detect

>> the holes of tape stripes.
>>
>> please see attached images to get an overview about my setup. in the other
>>
>> image I put some snapshots taken on a white silkscreen pcb, green
>> silkscreen pcb and at grey feeder. the glare at the feeders is removed
>> perfectly, but on the pcbs the fiducials are not good to be ssen.
>>
>> I would be interested if anybody had put polarizing filters into service
>> and how? Or am I on the wrong way here? What have you on your machines and
>>
>> how perfect works it for you? I think I would need some coaxial light?!
>>
>>
>>
>

> --
> You received this message because you are subscribed to the Google Groups
> "OpenPnP" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to openpnp+u...@googlegroups.com.
> To post to this group, send email to ope...@googlegroups.com.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/openpnp/9bb93cb6-167e-49f1-896b-f353fbeb7389%40googlegroups.com.

[Michael Gröne]

Am 19. September 2017 11:12:03 nachm. schrieb Cri S <phon...@gmail.com>:

> It seems that the camera calibration should be improved.
> Canny circle detection works without problems.
> If dimming and changing gamma (contrast) it should work.

Just not sitting at the machine. Ist gamma supported to change for the help
camera?

> For using it on loose feeder you should place 4 direct downlooking
> leds (normal leds)
> at the 4 edges in order to reduce vignetting, it is not optical, it's
> induced from lightning.

Yes, ist from lighting. As you can see in the posted stl file I have an
mount for downlooking ledstripe. That made a spot on left/right side below
the stripe but had no visible impact on pcbs in the view so I uncounted
these. But you're right, maybe just 4 leds in the corners would help.

> Eventually you should check if placing a baking paper at 30 degree
> over the leds helps
> for the center spot illumination drop. This after dimming module are installed.

Had ordered some diffuse foil. If I could get that illumination drop away,
that would be great.

> Example of fiducial preparation, with steps for visual checking, it
> works good for
> hough circle, don't know for template match, i see no reason that it
> don't work if adding
> masks, hovewer the holes near the fiducial can have better match on
> templatematch.

Oh yes, they have a better match. Had trouble with the old lighting setup
all way long.

Thank you for all your advice. I appreciate!

> You received this message because you are subscribed to a topic in the
> Google Groups "OpenPnP" group.
> To unsubscribe from this topic, visit
> https://groups.google.com/d/topic/openpnp/7yUivDAALIg/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to

> openpnp+u...@googlegroups.com.
> To post to this group, send email to ope...@googlegroups.com.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/openpnp/CAJGcfUh6C0He%3DRH%3D9ho-NbCKLZ_3iNYHCvmLX8BvORXkHm6SrQ%40mail.gmail.com.

[Michael Anton]

You should also back the solder mask off from the fiducials to increase the contrast, and so it is not dependant on the solder mask color.

[Michael Gröne]

Yeah, I used the fiducial offered in libs from sparkfun. It had no free area around it, except the standard clearance. Found an issue filed at github in the meantime at their lib repo.  It was first time designing using a fid and I didn't know that it was not optimal. Changed the design for next run already .

> --
> You received this message because you are subscribed to a topic in the Google Groups "OpenPnP" group.
> To unsubscribe from this topic, visit https://groups.google.com/d/topic/openpnp/7yUivDAALIg/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to openpnp+u...@googlegroups.com.

> To post to this group, send email to ope...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/872c2d44-e026-4481-81d6-438ff9f7103f%40googlegroups.com.

[Michael Anton]

The fiducials I use, are either a 40mil pad with an 80mil keepout, or a 100mil pad with a 200mil keepout.  You can pretty much use whatever pad size you want, but it is usually recommended to make the radius of the keepout beyond the pad, at least equivalent to the the radius of the pad.

[phon...@gmail.com]

40mil diameter with 80mil diameter keepout is the absolute minimal allowed specification.

Most imperial operating board assemblers have 0.050 +- 0.010 specs for fiducial and 0.100 for keepout.

Metric usually have 1-2mm +- 0.1mm  and 3R for Fiducial with  minimum 3mm for keepout and up to 1.5mm 3mm keepout is allowed/tollerated.

Example of good fiducial that allow backup strategy in case the tinning of fiducial create too mutch problems.

Tinning is mostly out of specs for smd fine pitch but used a lot for cost reasons.

https://www.google.it/imgres?imgurl=https%3A%2F%2Fwww.pcbuniverse.com%2Farticles%2FTooling%2520Rails%2C%2520Holes%2520and%2520Fiducials.png&imgrefurl=http%3A%2F%2Fwww.pcbuniverse.com%2Fpcbu-tech-tips.php%3Fa%3D1&docid=LYluxHvb4MnZXM&tbnid=x-FNP2f3hWppQM%3A&vet=10ahUKEwigxcePx7PWAhUJUhQKHaidAq8QMwgxKAswCw..i&w=1060&h=890&client=opera&bih=786&biw=1492&q=fiducials%20pcb&ved=0ahUKEwigxcePx7PWAhUJUhQKHaidAq8QMwgxKAswCw&iact=mrc&uact=8

[Michael G.]

tweaked lighting further. the histogram is more even now, changed from 6mm to an 8mm lens and widened the dome a little. just ordered a new one, the one I got from ELP is terrible (blur in the edges).
can't get the mirror of the camera away, still waiting for the dimming module. for reference some pictures attached.

[Cri S]

You have put the paper over the pcb, that gives false values and
don't permit you to check uniformity. If youi need to place it over pcb,
place twenty or more white paper over it, not just one.
The center of illumination using histogram should be on center of image,
actually is a lot out of center requiring cutting significant part of
image in order
to fix it using camera settings.
Further as this optics as i have checked it now with this image is not
rated for the sensor,
you should think about using the cropping 720p format instead of the
higher resolution
format.

> --
> You received this message because you are subscribed to the Google Groups
> "OpenPnP" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to openpnp+u...@googlegroups.com.
> To post to this group, send email to ope...@googlegroups.com.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/openpnp/3a9ded40-f37e-4f7c-8323-0c13567e99f8%40googlegroups.com.

[Cri S]

[Michael G.]

Hi Cri,

Am Donnerstag, 21. September 2017 00:34:08 UTC+2 schrieb Cri S:

You have put the paper over the pcb, that gives false values and
don't permit you to check uniformity. If youi need to place it over pcb,
place twenty or more white paper over it, not just one.

Will try this evening

 

The center of illumination using histogram should be on center of image,
actually is  a lot out of center requiring cutting significant part of
image in order
to fix it using camera settings.

I think its the lens why the bright center is off. that was not the case before changing from 6mm to 8mm. but I will look into that.

 

Further as this optics as i have checked it now with this image is not
rated for the sensor,
you should think about using the cropping 720p format instead of the
higher resolution
format.

How do you see in the pictures, that the lens is not rated for the sensor? I already got that feeling, but I thought its just a cheap one which is not so good. what irritates me is that its a lense bought in combination with the camera from the manufacturer directly. anyway, going to order this one now: https://www.amazon.de/gp/product/B006OIRM0I/ref=ox_sc_act_title_1?smid=A50DZI580G3JX&psc=1

 This one https://www.amazon.de/gp/product/B00REFQ116/ref=oh_aui_detailpage_o05_s01?ie=UTF8&psc=1 was working well as you have seen on the images posted last days.

Currently I'm at 1024x768 resolution. Thats pretty near to 720p and so okay, isn't it?

[Cri S]

On photoretouching program (gimp if you don't have) with color image
i have checked
achromatic correction doing channel separation to RGB and then disable
green layer
and enable/disable red layer in order to visually see the distortion.
On pcb, image i have normalised every channel image to see it better.
Beside checking the color distortion i saw that inside center the
image was sharp and
outside center it was optical blurred because the optics is rated only
for smaller chip sensors.

I know the pratice of selling cheap optics rated for two sensor sizes
smaller to high
pixel sensors. It could work if using it as webcam. It don't work for
image processing as
this relay on certain low level features of edge lines at certain blur
level that.
Check the sensor what you have, the resolution and crop /binning level
from datasheet
and select a more cropped resolution if you use this optic.

>> 2017-09-20 23:50 GMT+02:00, Michael G. <mic...@groene.de <javascript:>>:

>> > tweaked lighting further. the histogram is more even now, changed from
>> 6mm
>> > to an 8mm lens and widened the dome a little. just ordered a new one,
>> the
>> > one I got from ELP is terrible (blur in the edges).
>> > can't get the mirror of the camera away, still waiting for the dimming
>> > module. for reference some pictures attached.
>> >
>> >
>> > --
>> > You received this message because you are subscribed to the Google
>> Groups
>> > "OpenPnP" group.
>> > To unsubscribe from this group and stop receiving emails from it, send
>> an

>> > email to openpnp+u...@googlegroups.com <javascript:>.

>> > To post to this group, send email to ope...@googlegroups.com

>> <javascript:>.

>> > To view this discussion on the web visit
>> >
>> https://groups.google.com/d/msgid/openpnp/3a9ded40-f37e-4f7c-8323-0c13567e99f8%40googlegroups.com.
>>
>>
>> > For more options, visit https://groups.google.com/d/optout.
>> >
>>
>
> --
> You received this message because you are subscribed to the Google Groups
> "OpenPnP" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to openpnp+u...@googlegroups.com.
> To post to this group, send email to ope...@googlegroups.com.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/openpnp/54c915af-d836-47f0-b53b-812c8051c6bc%40googlegroups.com.

[Michael G.]

made some pictures to compare 2 lenses.

lens 1:
original one bought from ELP with camera
Imprint: 12mm 1/2.5" 5MP IR

lens 2:

Specs: 8mm ,1/3", F2.0,
Link: https://www.amazon.de/gp/product/B006OIRM0I/ref=oh_aui_detailpage_o00_s01?ie=UTF8&psc=1

camera:
ELP
Link: https://de.aliexpress.com/item/2mp-2-0-megapixel-1920-x-1080-cctv-camera-board/32263338732.html
Sensor: OV2710
Sensor Size: 1/3"


It seems, the 12mm lens is way better for my setup. Additionally I can see the same vignetting if having the dome installed or not. This makes me think its the combination of lens/camera is not optimal. Is the difference it makes the result better for the 12mm lens that its rated for sensor sizes of 1/2.5"? The vignetting area would be in the invisible area of the sensor, isn't it?

[Cri S]

lens2 8mm seems good. Can you provide orginal color image of white paper ?
The blue channel don't works. I don't know if this is the result of
bad focus or something
else.
For blur index,
https://www.pyimagesearch.com/2015/09/07/blur-detection-with-opencv/
Additionally split channels to bgr and do blur detection on separate
channel in order to
have 4 values, overall and separated for the channels.

>> 2017-09-21 9:05 GMT+02:00, Michael G. <mic...@groene.de <javascript:>>:

>> https://groups.google.com/d/msgid/openpnp/54c915af-d836-47f0-b53b-812c8051c6bc%40googlegroups.com.
>>
>>
>> > For more options, visit https://groups.google.com/d/optout.
>> >
>>
>
> --
> You received this message because you are subscribed to the Google Groups
> "OpenPnP" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to openpnp+u...@googlegroups.com.
> To post to this group, send email to ope...@googlegroups.com.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/openpnp/6fd18a65-d4bc-41e7-b52e-d714c630439a%40googlegroups.com.

[Michael G.]

hi cri,

attached a snapshot.
I don't see any problems with the blue channel. used imageplay to build a pipeline with the laplace-filter as in the blog post, see screenshot. Posted the .ipj file to load into imageplay, too. I don't see any blurry parts in the filtered blue channel.