Rotifers (fluorescence)

[emsiem]

Took photo of a slide with rotifera using EPI fluorescence.
Leitz Ploemopak, blue excitation, single shot.

Original photo came very bleached so I applied deconvolution (Huygens).
This again didn't look aesthetically pleasing to me so I tried overlaying original and deconvoluted image in PS (soft light).
Here are all three. I find overlaid, although lacking details of deconvoluted one, more pleasing.

Original photo


Deconvoluted photo


Overlaid

[Jacek]

nice

[Marek Mis]

Very interesting effect ! Nice images.

Marek

[RogelioMoreno]

Interesting!

Rogelio

[Charles Krebs]

Very nice! I too think the third version is the most pleasing... good idea!

[jswatts]

So I'm seeing a lot of reference to proprietary software for deconvolving single images on the forums lately. I use ImageJ for a lot of my image processing needs, and using that in conjunction with Photoshop was able to reproduce these results pretty well. If you're using software to "deconvolve" single images (rather than z-stacks) you're really just using a sharpening algorithm. There are all sorts of groovy things ImageJ is capable of, it's used by professionals in many fields all over the world, and it's FREE! Well, I mean, it came out of the US taxpayers' taxes, but what a bargain! Hopefully this will be helpful to the inspired amateurs producing these stunning images.

The last pic of the three is definitely the most effective. Great work!

[Pau]

Nenad,
just as an exercise I've reworked your first image in PS with heavy use of levels and unsharp mask and the result is very similar to your last image, may be a bit better for my taste.

If you agree I may post it for comparison

[emsiem]

Thanks to everyone for comments.
Pau, I would like to see your results. I usualy also use PS postprocessing, but compared to deconvolution most often there is less details.

[emsiem]

So I'm seeing a lot of reference to proprietary software for deconvolving single images on the forums lately. I use ImageJ for a lot of my image processing needs, and using that in conjunction with Photoshop was able to reproduce these results pretty well. If you're using software to "deconvolve" single images (rather than z-stacks) you're really just using a sharpening algorithm. There are all sorts of groovy things ImageJ is capable of, it's used by professionals in many fields all over the world, and it's FREE! Well, I mean, it came out of the US taxpayers' taxes, but what a bargain! Hopefully this will be helpful to the inspired amateurs producing these stunning images.

The last pic of the three is definitely the most effective. Great work!

Thank you for comment.

Deconvolution is in fact different from standard sharpening methods. It is itterative algorithm with goal to approximately reverse bluring proces using calculated PSF (meassured or theoretical). Sharpening methods dont care about how photo was taken so more information is lost.

I tried ImageJ deconvolution plugins, but it lacks option to calculate theoretical PSF and I don't have nano fluorescent particles to create meassured PSF. Maybe there is some plugin to get parameter calculated PSF for ImageJ?

[Pau]

Here you have



I've played quite strongly with levels, brightness and contrast, selective correction increasing blacks and a bit greens and fairly strong unsharp mask.

Playing with original 16 bit images and with more careful processing it may end better.

[emsiem]

Hey Pau, that looks really good!
Of course, you worked on already reduced JPEG image (800x685) and deconvolution/overlaying was applied on TIFF (1600x1369) so I imagine your results could have been even sharper.

If you can see, my 3rd image have little bit more details in green channel, but I haven't initially provide enough details on the process to explain why.
I used only green channel deconvolution results in grayscale mode as layer for sharpening (using soft light overlay mode) so that is the reason red and yellow areas are much more blurred.
What deconvolution do really well is background detection - subtraction, which is very hard in PS on some subjects.

Anyway, I like your results and it's hard to tell for me which one is better
Thanks for sharing your technique.

[jswatts]

So I'm seeing a lot of reference to proprietary software for deconvolving single images on the forums lately. I use ImageJ for a lot of my image processing needs, and using that in conjunction with Photoshop was able to reproduce these results pretty well. If you're using software to "deconvolve" single images (rather than z-stacks) you're really just using a sharpening algorithm. There are all sorts of groovy things ImageJ is capable of, it's used by professionals in many fields all over the world, and it's FREE! Well, I mean, it came out of the US taxpayers' taxes, but what a bargain! Hopefully this will be helpful to the inspired amateurs producing these stunning images.

The last pic of the three is definitely the most effective. Great work!

Thank you for comment.

Deconvolution is in fact different from standard sharpening methods. It is itterative algorithm with goal to approximately reverse bluring proces using calculated PSF (meassured or theoretical). Sharpening methods dont care about how photo was taken so more information is lost.

I tried ImageJ deconvolution plugins, but it lacks option to calculate theoretical PSF and I don't have nano fluorescent particles to create meassured PSF. Maybe there is some plugin to get parameter calculated PSF for ImageJ?

Hi, emisem,

There's deconvolution and there's deconvolution. If you have a full 3D data set then the iterative process you describe can be applied. These methods require that both the in-focus and out-of-focus light from all parts of your sample be acquired. They then use mathematical methods the put the out-of-focus light back where it belongs. Some require the PSF to be measured beforehand (non-blind) while others do not (blind).

However, if you have but a single image (and you've indicated that your lovely photo is a single shot), then you can't apply these techniques and all you have available are basically sharpening methods. Since you don't know where it came from, you can't reassign the out-of-focus light back to its origin. The only similarity between full 3D deconvolution and 2D deconvolution is that they both deal with the Fourier transform of the image.

When I had a play with your image, I just used background subtraction in ImageJ to remove the uneven background fluorescence in your image. The I too it into Photoshop, duplicated the image as a new layer and ran a high pass filter on it. The filtered image layer was then set to soft light blending mode to achieve the sharpening. I hope this helps.

[emsiem]

Hello

Yes, you are right, there is deconvolution and deconvolution as deconvolution is only a name for concept of reversing convolution and multiple procedures and algorithms are used for different approaches and inputs.
But saying that 3D deconvolution is only real one is not correct.
Deconvolution is very often used in astronomy and there you can't have 3D data sets.
It is also used for removal of noise in sound or other 1D signal (hence the name 1D deconvolution)

Sharpening is image enhancement technique and deconvolution is image restoration technique.

2D deconvolution methods try to reconstruct convolution process based on known properties of optical system (i.e. objective NA, aperture, mounting medium refraction index,...) and statistical analysis of image to be restored which as result gives calculated PSF. That PSF is then applied in reverse process (deconvolution) to remove noise, optical errors and out of focus parts of image.

Sharpening methods (i.e. unsharp mask) only try to amplify difference in brightness on edges in the image.

That is why 2D deconvolution is not sharpening method.

Nevertheless, my goal was to visually enhance the image so true, image enhancement methods could have been used, but I wanted to experiment.

Thank you for debate :-)