Up to now the only decent dots I've been able to resolve were on a Carter slight lent me by James. On other slides, René's for example, I could resolve dots but not over the length of a valve or in rows of more than three or four.

Recently I've made some strews myself of KK material and have used RegiStax 4.0 on some AVI files (640 X 480). I have posted a picture that demonstrates that while it is possible to see them, it's not easy to make a decent picture:
  Link : http://www.science-info.net/pages/EDFWilliams/PICTURE/758dots.jpg

This was made from 647 frames (97 stacked) taken with oblique illumination from an intermediate lens. A normal blue filter was used. Obj: Lomo Apo 90/1.3; Ocular: Periplan 12.5/18 Photo; Condenser: Lomo Aplanatic 1.4 with iris wide open. The intermediate lens used to tilt the beam was from a scrapped flatbed scanner. Immersion oil (Zeiss 515N) on condenser and objective.

Because the Lomo frame I use has Epi as well as transmitted lighting with Epi having almost twice the light path length of transmitted, the filament image can be focussed over a very large range and the focal length of the intermediate lens hardly matters. I've used three different lenses and they all work well. Furthermore the modified lamp holder itself can be moved backwards more than an inch.

Don W

There is a video of the same A pellucida specimen here:
  http://www.science-info.net/pages/EDFWilliams/VIDEO/760DX.avi

It might be a good idea to lean back and slit your eyes, or look through a glass darkly to see the rows of dots go in and out of focus. Imagination would help too.

Don W

I came to the conclusion some time ago that the Diffraction Contrast method was nothing more than oblique lighting.

To check on this belief I set up an edge (Convex) in the filter holder of one of the Aplanatic Condensers. I adjusted the 'knife-edge' until the lines were visible on the specimen -- and then with everything else the same as before -- I took some 850 video frames. I stacked them with Registax and fiddled with the 6:1 Wavelet Slider a bit.
The result is as expected:
  Link : http://www.science-info.net/pages/EDFWilliams/PICTURE/772REG353.jpg
By the way Registax decided to stack 353 frames.

The only difference between an offset condenser iris and the 'diffraction knife edge' was that the knife edge is convex (44mm radius) and the iris is concave (about 18mm radius). They are at very slightly different heights as well.

Now I'm going to cover the microscope and have a glass of wine and some dinner.

Don W

Hello Don,
> ...it's not easy to make a decent picture:
> http://www.science-info.net/pages/EDFWilliams/PICTURE/758dots.jpg

This picture looks great.

Did you make that with your Logitech QuickCam? I think that has a CMOS chip. My experience with CCD-camera's is that they perform better (less noise). Did you compare these different types of camera?

Kind regards, Marien

Yes it's the Logitech. I have a couple of other cameras here now. I know nothing about them except they come from China. As soon as I have an adapter I can try one of them and see what happens.

In the meantime I have used Micam to record 604 frames, but this time saved as Div-X compressed. Then I used Registax again. But it took the program more than an hour to get through the file and the result is more or less the same as before.

Don W

Hi Don,
Another interesting look at your new K.Kemp A.pellucida strew.

It doesn't take too much imagination to discern "the dots" as shown here:
  Link : www.kscitech.com/MicGroup/Temp/DW_01.htm

Thanks again for sharing your pics with us.
-Keith

"Don Williams" wrote:
> There is a video of the same A pellucida specimen here:
> http://www.science-info.net/pages/EDFWilliams/VIDEO/760DX.avi
> ...see the rows of dots go in and out of focus.
> Imagination would help too.

Hi Don
> Yes it's the Logitech.

Do you cool the camera? that will keep the noise down, and what shutter speed?

David

Hello Marien,
Late last year I was evaluating some high-resolution webcams (HRWC) as described here:
  Link : www.kscitech.com/Webcam

This webpage includes some links to discussions on CCD sensors and the transition to CMOS sensors, as promoted by Micron.

Don uses the Logitech Pro 4000, as listed in this comparison chart.

I also had some comparison information re the Toucam and Vesta cameras, but can't find it right now....

I hope this summary may be helpful.

Regards, Keith

"marienvanwesten" wrote:
> Did you compare these different types of camera?

Using the same optical arrangement as in the previous few posts I set about looking for dots with a wide open condenser diaphragm and no intermediate lens to deflect the cone of light. Instead I used the same concave knife edge in the filter holder of the condenser and employed the rotation and shift to place the knife edge in the best position to show lines or dots. The first two are stacks (Registax 4) of 108 and 21 frames respectively and dots can be seen. In the last picture the knife edge was set to give the most contrast and show the lines as clearly as possible, here Registax chose to stack 342 frames.

Images (640x480):

  Link : www.science-info.net/pages/EDFWilliams/PICTURE/773REG108.jpg

  Link : www.science-info.net/pages/EDFWilliams/PICTURE/774REG21773.jpg

  Link : www.science-info.net/pages/EDFWilliams/PICTURE/775REG342.jpg

It would seem that oblique lighting achieved by the use of a knife edge is just as good as using an intermediate lens, or shifting the condenser diaphragm. I would also tentatively suggest that it is better for this specimen than COL. I have left out a lot of detail, but previous posts will provide information about the microscope and camera.

I am posting these pictures as I go along rather than filing them away where they could soon be forgotten amongst the thousands of others accumulated over the past five or six years.

Don W

Hi Don,
> I have left out a lot of detail, but previous posts
> will provide information about the microscope and camera.
> I am posting these pictures as I go along rather than filing them
> away where they could soon be forgotten amongst the thousands of
> others accumulated over the past five or six years.

I uploaded a Topic summary page:
  http://www.kscitech.com/MicGroup/Msg_08.1.htm
which may assist members in following this work.
It includes a comparison of the initial oblique and DLC technique images:
  http://www.kscitech.com/MicGroup/DW_02.htm

> ...concave knife edge in the filter holder of the condenser and
> employed the rotation and shift to place the knife edge in the
> best position to show lines or dots.

This is definitely a superior mechanical arrangement. I understand that you are using the Lomo oblique condenser with the ON-14 designation. Where is the filter holder located in relation to the condenser lens and the iris?

Best regards, Keith

Hi,
A query about your use of Registax please. Are you combining images in order to average out noise from a low resolution video and/or are you combining images of differing focus? I can't find a description of Registax on their site but I am presuming that this programme has similar functionality to an ImageJ ad-on that lines up a stack of images after which one can take an average with no other processing done.

Whatever, your images are impressive.

If the subject is static during the video take then presumably the light information always maps to the same pixels and thus apparent video resolution (other than noise) cannot be enhanced. Are you moving the slide slightly and relying on Registax to re-align the images?

What software do you use to break a video into frames?

Selwyn

Hi Selwyn,
You wrote:
> Are you combining images in order to average out noise from a low
> resolution video and/or are you combining images of differing focus?

You bring up a good point re 'averaging' or 'stacking'.
Normally, 'stacking' refers to a sequence of images that are processed to enhance depth of field (DOF) i.e. to produce a pseudo-3D output image. The algorithms used involve some sophisticated re-alignment of input images.

My understanding is that Marien's Ap image was produced from a simple 'averaging' of static images, without re-alignment, with his MICAM software.

Now both Peter and Don refer to 'stacking' to produce their "best" Ap images. As far as I can ascertain Peter does not elaborate on the image processing used to obtain the Ap image he posted with this group, but Don is explicitly using Registax. Furthermore, I found a note that Registax "uses a Fourier transformation for initial image alignment". I'm not sure why this is required with a sequence of static images, such as those from an Ap slide, acquired using a video camera/webcam. Unless there is some inherent "jitter" frame-to- frame?

Hopefully Peter and Don will help us understand exactly what image processing techniques are being used.

> What software do you use to break a video into frames?

Evidently Registax runs with a video file as input. You will note that Don mentions Registax "dropping" frames during processing.

The freeware VirtualDub program does a decent job of manipulating a video file frame-by-frame.

Hope this helps some.

Regards, Keith

K3CCDTools software will do it also

David

Hello to all,
I will try to explain, how I receive my A.p. pictures.

First is that I use a camera with a 12bit resolution in 640 x 480 pixel. Its normally for astronomy. The type is Lumenera SKYnyx 2-0M. Its a monochrome type. I have had no success with a normal webcam!

I use the software K3CCD for the recording. Between 500 to 1000 frames in an AVI video are made.
The first AVI is the picture WITH the A.p.!
Next I made an AVI with the same paramerers, but WITHOUT the A.p. in the picture. Only to have the illumination , the noise , the demosaicing or other troubles in the picture.

Next I process the 2 AVIs with the same parameters in Registax 3. And I use only the point "create a flatfield". This is add together all frames, nothing more.

After this step I have 2 pictures. One with the A.p. and all the noise, second picture only the noise.

Now I take the program GIOTTO(freeware). There I made the correction, adjustment. The result is a low-noise picture, without the trouble of the illuminating and so on..

Next important point is the use of the UV-LED near 385nm! Only with this wavelength I had the success to get high-contrast pictures.

I hope I got you some important points.

Best regards, Peter

Hello Peter,
You wrote:
> I hope I got you some important points.

Yes indeed - very much appreciated!
It's dinner time right now in Massachusetts, but we'll follow up in due course.

Best regards, Keith

"scitech200" wrote:
> My understanding is that Marien's Ap image was a simple 'averaging'
> of static images, without re-alignment, with his MICAM software.

You are right Keith, In the experimental version of MICAM I use now I can average four consecutive captured frames (the only reason to do that is minimizing noise). I am working on a version with the ability to average more frames, to see were the optimum is. I think the optimum will be at 8 or 16 frames, Don also said averaging many frames doesn't give a much better result.

I wrote MICAM because I wanted to save a captured frame with additional information, just with one mouse click. I also want to incorporate features from my NAFIS program, were you can select a name from a database to give the image a meaningful name. That really saves a lot of typing.

When ready I will put this version with the averaging capabilities on my site.

Kind regards, Marien

Some members may not be aware that there is a simple statistical calculation for the reduction in noise achieved by averaging (arithmetic mean) frames. Taking the standard deviation as a measure of the spread of noise across one pixel in multiple frames the standard deviation of that pixel, in the average of N frames, is reduced to 1 divided by the square root of N of its original value, showing that as N increases pixel values get increasingly closer, on average, to the long term mean.

Thus the average of 4 images can be said to halve the noise, the average of 100 images to reduce it to a tenth, etc. It is apparent that diminishing returns soon set in.

Selwyn

For the sake of purists I add that the rule relating sampling error to the size of sample is independent of assumptions about the nature of the underlying statistical distribtion, with one exception: the Cauchy-Lorentz distribution. In that pathological case, which cannot arise in this context because the range is finite and the distribution discrete, the image from medians would be required. One could always take the median in preference to the mean but this is statistically less efficient when the general rule applies.

Selwyn

Hi,
I really haven't been following this thread with much interest; from a distance, it seems to be another "all-previous-science-is-wrong-I've-broken-the-Abbe-Limit" themes. Maybe that's not a fair assessment, but I've heard the boy cry "wolf" too many times. However, for those who are really interested in this, have you tried comparing DLC with Hoffman Modulation Contrast?

http://micro.magnet.fsu.edu/primer/techniques/hoffman/hoffmanindex.html

I'm not making any claims here, and I may even be off-base. But I just wanted to introduce some critical thinking into the mix. It seems like Mr. Piekos has re-invented the wheel, or at best put some new racing hubs on the old one.

Don H