Compare Images to Quantify Cleaning Performance

[Frank Robson]

Hello,

I'm trying to quantify the cleaning performance of a dishwasher. To do so, I have standardized "dirt test plates" that I clean using the machine.
I take pictures of the test plates before and after cleaning them and want to measure how much dirt got off during the process.
I thougth ImageJ could help me out to do so but the only metric I found for comparision until now are the RGB-values of both images.
I was wondering if there was a better way of doing it.

After the wash, the plate becomes brighter and the stains more grainy.
Pixel to pixel comparision is not possible because the camera is not held still.

Thank you for your help,
Frank

[gankaku]

Hi Frank,

could you provide an example image for the before and after comparison,
respectively.
Depending on how the differences look like, we might be able to suggest a
possible solution.

In general, without having seen the images, I would suggest that a standard
imaging setup makes sense. Thus, I would try to always keep the exact same
distance and relative position to the plate.
In the worst case construct yourself a scaffold for the plates with a fixed
distance to the camera tripod.

regards,
Jan

2015-09-30 17:18 GMT+02:00 Frank Robson <[hidden email]>:

> Hello,
>
> I'm trying to quantify the cleaning performance of a dishwasher. To do so,
> I
> have standardized "dirt test plates" that I clean using the machine.
> I take pictures of the test plates before and after cleaning them and want
> to measure how much dirt got off during the process.
> I thougth ImageJ could help me out to do so but the only metric I found for
> comparision until now are the RGB-values of both images.
> I was wondering if there was a better way of doing it.
>
> After the wash, the plate becomes brighter and the stains more grainy.
> Pixel to pixel comparision is not possible because the camera is not held
> still.
>
> Thank you for your help,
> Frank
>
>
>
> --
> View this message in context:
> http://imagej.1557.x6.nabble.com/Compare-Images-to-Quantify-Cleaning-Performance-tp5014503.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--

CEO: Dr. rer. nat. Jan Brocher
phone:  [phone]
mobile: [phone]
e-mail: [hidden email]
info: [hidden email]
inquiries: [hidden email]
web: www.biovoxxel.de

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Mike Esterman]

Frank
You might try subtracting the clean plate image  from the dirty plate image
the resulting image will be the dirt - threshold the difference image but
don't apply that will select any particles, or if you want the entire plate
set the threshold to select the entire difference image and could the image
by calculating the mean density of the image.

Hope this helps

Mike

-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Frank
Robson
Sent: Wednesday, September 30, 2015 11:18 AM
To: [hidden email]
Subject: Compare Images to Quantify Cleaning Performance

Hello,

I'm trying to quantify the cleaning performance of a dishwasher. To do so, I
have standardized "dirt test plates" that I clean using the machine.
I take pictures of the test plates before and after cleaning them and want
to measure how much dirt got off during the process.
I thougth ImageJ could help me out to do so but the only metric I found for
comparision until now are the RGB-values of both images.
I was wondering if there was a better way of doing it.

After the wash, the plate becomes brighter and the stains more grainy.
Pixel to pixel comparision is not possible because the camera is not held
still.

Thank you for your help,
Frank



--
View this message in context:
http://imagej.1557.x6.nabble.com/Compare-Images-to-Quantify-Cleaning-Perform
ance-tp5014503.html
Sent from the ImageJ mailing list archive at Nabble.com.

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Frank Robson]

Hello everyone,

thank you for the answers.
Here are some sample images of the test plates before and after the cleaning process:


The distance between plate and camera is always kept the same, however the angle might differ. I built a kind of blackbox so that no external lighting will infer (the camera is always kept in manual mode).
I'll have a look if it would be possible to use a tripod in order to assure a constant angle for the camera.

Cheers,
Aaron

[Peter Haub]

Hi Frank,

ensure that the images have identical scale and align both images via
phase correlation. Then it should be possible to detect 'the' image
differences in the overlapping region of both images by 'Image
Calculator/ Difference'. ('The' differences means at least areas where
objects are vanished etc.  )
Take care of image intensity, color and contrast. Image differences on
that level should not originate from unstable illumination conditions or
varying camera settings.

In case of different image scales you have to map the images in relation
to reference points (which can be either selected manually or found
automatically e.g. by techniques like SIFT).

Regards
Peter

On 01.10.2015 10:11, Frank Robson wrote:

> Hello everyone,
>
> thank you for the answers.
> Here are some sample images of the test plates before and after the cleaning
> process:
> <http://imagej.1557.x6.nabble.com/file/n5014514/V6vz.jpg>
> <http://imagej.1557.x6.nabble.com/file/n5014514/V6n1z.jpg>
> The distance between plate and camera is always kept the same, however the
> angle might differ. I built a kind of blackbox so that no external lighting
> will infer (the camera is always kept in manual mode).
> I'll have a look if it would be possible to use a tripod in order to assure
> a constant angle for the camera.
>
> Cheers,
> Aaron
>
>
>
> --
> View this message in context: http://imagej.1557.x6.nabble.com/Compare-Images-to-Quantify-Cleaning-Performance-tp5014503p5014514.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[gankaku]

Hi Frank,

joining the opinion and statement of Peter, I would also suggest to try to
improve imaging as much as possible for equal conditions.
Having said this, the following macro does the job quite well. I would run
it in Fiji since you need the Linear Stack alignment with SIFT plugin from
Stephan Saalfeld, thus using SIFT as suggested by Peter which is installed
by default in Fiji.
It aligns the images scale and orientation wise. Note, this affects pixels
to some extend. So, I would be careful afterwards with absolute pixel
intensity or color difference comparison (avtually I wouldn't do it).

Nevertheless, this method enables you to see where you have objects left or
washed off the plate. The difference will show you changes between the
plates. It does not distinguish if something was present first and then
washed off or was not present but appeared after treatment (which I guess
in your case is rather unlikely).

How to:

open the two images you want to compare (and only those two since the macro
is not adjusted to run on specifically selected images so far)
Copy the macro in Fijis' scripting editor (>Plugins >New >Macro) and hit
the run button.

//macro start ------------------------

run("Images to Stack", "method=[Copy (center)] name=Stack title=[]");
run("Linear Stack Alignment with SIFT", "initial_gaussian_blur=1.60
steps_per_scale_octave=3 minimum_image_size=64 maximum_image_size=1024
feature_descriptor_size=4 feature_descriptor_orientation_bins=8
closest/next_closest_ratio=0.92 maximal_alignment_error=1 inlier_ratio=0.05
expected_transformation=Affine interpolate");
run("Duplicate...", "duplicate");
run("Stack to Images");
imageCalculator("Difference", "Aligned-0001","Aligned-0002");
close("Aligned-0002");

//macro end ------------------------



Hope this helps you a little further.
Kind regards,
Jan

2015-10-01 10:11 GMT+02:00 Frank Robson <[hidden email]>:

> Hello everyone,
>
> thank you for the answers.
> Here are some sample images of the test plates before and after the
> cleaning
> process:
> <http://imagej.1557.x6.nabble.com/file/n5014514/V6vz.jpg>
> <http://imagej.1557.x6.nabble.com/file/n5014514/V6n1z.jpg>
> The distance between plate and camera is always kept the same, however the
> angle might differ. I built a kind of blackbox so that no external lighting
> will infer (the camera is always kept in manual mode).
> I'll have a look if it would be possible to use a tripod in order to assure
> a constant angle for the camera.
>
> Cheers,
> Aaron
>
>
>
> --
> View this message in context:
> http://imagej.1557.x6.nabble.com/Compare-Images-to-Quantify-Cleaning-Performance-tp5014503p5014514.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--

CEO: Dr. rer. nat. Jan Brocher
phone:  [phone]
mobile: [phone]
e-mail: [hidden email]
info: [hidden email]
inquiries: [hidden email]
web: www.biovoxxel.de

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Frank Robson]

Hello,
Thank you both for the very insightfull comments ! I tried out the macro and it produced very interesting results. It seems a very promising way to proceed with this type of test plates.
However I got also a different type of these, where the washing doesn't spill off particles but rather bleaches the color of the dirt. Here are some sample photographs :
Before:
After:
I guess the previous way of doing wouldn't work as well on these... This is why I initially thought of using the RGB coloration values. Would you know a way to proceed in this case ?

Thank you again for the great help.
Best,
Frank

[Peter Haub]

Hi Frank,
I would not spend time thinking about analyzing those images. Invest
some time to improve the image quality. This will for sure simplify the
analysis and enhance the quality of the measurement.
Regards,
Peter


On 01.10.2015 15:14, Frank Robson wrote:

> Hello,
> Thank you both for the very insightfull comments ! I tried out the macro and
> it produced very interesting results. It seems a very promising way to
> proceed with this type of test plates.
> However I got also a different type of these, where the washing doesn't
> spill off particles but rather bleaches the color of the dirt. Here are some
> sample photographs :
> Before:  <http://imagej.1557.x6.nabble.com/file/n5014518/V7v_z.jpg>
> After:  <http://imagej.1557.x6.nabble.com/file/n5014518/V7n_z.jpg>
> I guess the previous way of doing wouldn't work as well on these... This is
> why I initially thought of using the RGB coloration values. Would you know a
> way to proceed in this case ?
>
> Thank you again for the great help.
> Best,
> Frank
>
>
>
> --
> View this message in context: http://imagej.1557.x6.nabble.com/Compare-Images-to-Quantify-Cleaning-Performance-tp5014503p5014518.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html