*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****
Dmitry,
This list is not about imaging faces, it is about confocal microscopy. We have one criterion only, the Rayleigh resolution. I advise students to, where possible, acquire images with 3 pixels within the Rayleigh resolution of their objective. This is slight oversampling, but it is justified for two reasons: (1) it makes details at the resolution limit easier to see, and (2) it allows for mild (VERY mild) filtering to remove noise without affecting resolution. Any further oversampling just bleaches the specimen without giving any further information. So now we have acquired our data. We are scientists, not photographers, so now we have two things to do, ANALYSE our data (which can involve a whole string of operations) and PRESENT our data. In the presentation phase we cannot expect a journal to magically give its readers the value of each pixel on the printed page - the image is printed with a halftone screen, for a start, and also it will have a particular gamma so that the original data is not reproduced linearly (actually your computer screen doesn't have a linear gamma either, though high-end imaging programs will allow to adjust this).
If we print a 35mm negative for publication, we'll enlarge it to 3" wide for a 1-column image or 7" wide for a full-page presentation. We'll also pick a grade of paper which gives the correct contrast for the purpose. Working with digital images we should do exactly the same. This is not changing the data (which I hope is safely stored somewhere) it is presenting it. Likewise, when I analyse intra-membrane particles I don't present every measurement, I publish a histogram, with mean values and standard deviations. But of course I still have the measurement of every particle. The reader does not expect to see the raw data - in the case of a digital image on paper that is impossible, and in the case of a particle analysis it would be incomprehensible - he or she expects to be presented with what the data reveals.
Guy
Optical Imaging Techniques in Cell Biology
by Guy Cox CRC Press / Taylor & Francis
http://www.guycox.com/optical.htm
______________________________________________
Associate Professor Guy Cox, MA, DPhil(Oxon)
Australian Centre for Microscopy & Microanalysis,
Madsen Building F09, University of Sydney, NSW 2006
Phone +61 2 9351 3176 Fax +61 2 9351 7682
Mobile 0413 281 861
______________________________________________
http://www.guycox.net
-----Original Message-----
From: Confocal Microscopy List [mailto:[log in to unmask]] On Behalf Of Dmitry Sokolov
Sent: Friday, 9 September 2011 10:21 PM
To: [log in to unmask]
Subject: Re: Preparing figures for publication --PPI vs DPI
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****
Dear All,
I am afraid that the confusion about Nyquist sampling is still there.
Nyquist-Shannon theorem is about full digital description of analogue
signal with highest possible frequency. This is conventionally applied in
its pure form to critical dimension (CD) imaging where the techniques are
facing their physical limitations.
The same theorem applied to non-CD imaging and perception of digital
images refers to number of pixels per feature of interest rather than
resel. Features of interest in optical imaging of a human face and
electron imaging of a letter on a coin for example will have negligibly
low spatial frequency compared to theoretical. The features will be
successfully resolved when presented by more than 2 pixels in the image. I
usually suggest 2-10 pixels per a feature as the lower limit to my
students. Pixels in this case are either physical pixels on the screen or
a resels on paper (0.1mm). The actual number of pixels per feature depends
on the size of the overall structure to be presented in the image. A
feature as small as 1/1000 of the width of the image will require zooming
to be perceived or resolved by the human eye.
The same magic number 1000 is involved in printing and computer interface
matters. 3" 300 dpi image will contain about a 1000 pixels across.
Conventional displays and data projectors have slightly more than a 1000
pixels along the screen. You may be probably agree now that 1000 pixels
images are the current standard for scanning and digital imaging. Bigger
number of pixels will create slightly better looking images according to
Nyquist criteria on the expense of higher acquisition time and lower
intensity of signal according to "Triangle of Frustration" principles:
http://confocal-manawatu.pbworks.com/w/page/37083554/Triangle%20of%20Frustration
I believe that 300 dpi is a big misleading requirement to the images. 3"
images in the text and 7" image on a cover of a journal will have
different number of pixels at the same pixel density. PPI at an image
printed size advised as a reference is what I would vote for.
Concerning the resampling of pixelated images, it's no nice at all. In my
old student time no data would be allowed without explanation and
validation of all of the processing steps and procedures. Pixel to resel
or pixel to pixel ratio should be 1x1 in ideal case for any kind of
presentation. Publishing on a cover would require high-resolution primary
data to remain scientific. Those are the ideal requirements. The actual
situation with presentation and publishing of visual scientific data is
however drastically different.
Here is the permalink at MIAWiki on the topic to those who is interested
in developing the "final" Wiki page:
http://confocal-manawatu.pbworks.com/w/page/45307622/Nyquist%20Sampling%20for%20non-CD%20Imaging
Thank you,
Dmitry
On Fri, September 9, 2011 9:27 pm, Guy Cox wrote:
> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Mark,
>
> Journals generally want the images to be 300dpi at their final
> reproduction size. I agree that doesn't entirely make sense but even
> as the editor of a microscopy journal I can't persuade production to
> see things any other way. So if you just change the declared
> resolution in the header to 300 dpi on a 512x512 image it will come
> out very small and you'll see nothing. As to your case, if you have
> checked the TIFF header and it says 300dpi I can't quite see how
> anyone could claim it's 258 dpi! However, I cannot see any merit in
> seeing individual pixels except in very special cases. As I said in a
> previous post - and as you know very well - the Nyquist criterion is
> based on the highest sampling rate being reproduced as a sine wave.
> If it appears as blocks (square waves) there are axiomatically higher
> frequencies present.
>
> We have to consider the human visual system, too. This has a built in
> and very powerful edge-recognition algorithm. This in turn means that if
> pixels are visible to actual data often is not.
>
> Guy
>
> Optical Imaging Techniques in Cell Biology
> by Guy Cox CRC Press / Taylor & Francis
> http://www.guycox.com/optical.htm
> ______________________________________________
> Associate Professor Guy Cox, MA, DPhil(Oxon)
> Australian Centre for Microscopy & Microanalysis,
> Madsen Building F09, University of Sydney, NSW 2006
>
> Phone +61 2 9351 3176 Fax +61 2 9351 7682
> Mobile 0413 281 861
> ______________________________________________
> http://www.guycox.net
>
>
>
> -----Original Message-----
> From: Confocal Microscopy List [mailto:[log in to unmask]]
> On Behalf Of Mark Cannell
> Sent: Friday, 9 September 2011 7:08 PM
> To: [log in to unmask]
> Subject: Re: Preparing figures for publication --PPI vs DPI
>
> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Hi Guy
>
> I don't think that's quite right, surely it's the final production
> magnification that determines whether the original structures are resolved
> at 300 DPI?
>
> FWIW, just last week had a Wiley production compartment (in China!)
> complain that one of my pics was not 300 DPI (they said it was 258 DPI !).
> The funny thing is that the images were all 300 DPI when I sent them off
> (would not have passed their brainless image quality check tool
> otherwise). They didn't like the new copy I sent either (still 300 DPI)
> and asked me to send them the original data so they could process it
> correctly for me! The issue was almost certainly that they did not like
> the fact that you could actually see individual pixels in the image. I on
> the other hand thought it quite acceptable -but what do I know about it...
> Since the processing of the image took about many many lines of IDL code I
> politely declined their kind offer to reprocess.
>
> Cheers Mark
>
> On 9/09/2011, at 9:28 AM, Guy Cox wrote:
>
>> *****
>> To join, leave or search the confocal microscopy listserv, go to:
>> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>> *****
>>
>> Brian,
>>
>> In my view you are indeed missing something. Nyquist sampling
>> is ~2.3 pixels per resel. At 300dpi these pixels would be 85µm
>> in size, so you couldn't see structures that are resolved in the
>> image. A 512x512 image would be 43 mm or 1.7" square.
>>
>> The aim of publication is communication, so to communicate your
>> results you need to present them in a way that they can be
>> understood when published in a journal. In any case, a confocal
>> image is just a series of sample points - Nyquist theory assumes
>> that these will be presented as sine waves, not blocks. So mapping
>> your samples into a higher-resolution space is in fact the correct
>> thing to do. By all means put your original data into an archive,
>> and I wish more journals would provide this facility (though equally
>> I can understand why they don't).
>>
>> If you are my age (or even half my age) you would once have recorded
>> your images on 35mm film. Did you insist on publishing them at that
>> size (24x36mm)? Of course not, you printed the negatives at the
>> required size. That's changing the data a LOT more than bicubic
>> resampling.
>>
>> Guy
>>
>> Optical Imaging Techniques in Cell Biology
>> by Guy Cox CRC Press / Taylor & Francis
>> http://www.guycox.com/optical.htm
>> ______________________________________________
>> Associate Professor Guy Cox, MA, DPhil(Oxon)
>> Australian Centre for Microscopy & Microanalysis,
>> Madsen Building F09, University of Sydney, NSW 2006
>>
>> Phone +61 2 9351 3176 Fax +61 2 9351 7682
>> Mobile 0413 281 861
>> ______________________________________________
>> http://www.guycox.net
>>
>>
>>
>> -----Original Message-----
>> From: Confocal Microscopy List [mailto:[log in to unmask]]
>> On Behalf Of Armstrong, Brian
>> Sent: Friday, 9 September 2011 3:32 AM
>> To: [log in to unmask]
>> Subject: Re: Preparing figures for publication --PPI vs DPI
>>
>> *****
>> To join, leave or search the confocal microscopy listserv, go to:
>> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>> *****
>>
>> Is the request to upload an image that has increased Dpi?
>> For this I would open the image in Photoshop, open image size dialogue
>> box, type in resolution 300dpi, UNCHECK Resample image, and choose OK.
>> It seems to me that resampling a scientific image using an algorithm
>> such as bicubic is interpolating pixels and therefore creating new data
>> where it did not exist before.
>> Am I missing something?
>>
>> Brian Armstrong PhD
>> Assistant Research Professor
>> Light Microscopy Core
>> Beckman Research Institute
>> City of Hope
>> 1500 East Duarte Road
>> Duarte, CA 91010
>> 626-256-4673 x62872
>>
>> Light Microscopy Core Facility
>>
>> -----Original Message-----
>> From: Confocal Microscopy List [mailto:[log in to unmask]]
>> On Behalf Of Guy Cox
>> Sent: Thursday, September 08, 2011 5:15 AM
>> To: [log in to unmask]
>> Subject: Re: Preparing figures for publication --PPI vs DPI
>>
>> *****
>> To join, leave or search the confocal microscopy listserv, go to:
>> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>> *****
>>
>> David,
>>
>> I suppose it's partly Photoshop trying to guess what users
>> want, and partly a matter of what's computationally possible. If you
>> are talking about 8-bit indexed color (ie grayscale with a palette
>> added) there is no way you can you can apply any of the sophisticated
>> resampling algorithms to it. Turning it into RGB and resampling is
>> indeed wrong, very wrong, what you must do is turn it into grayscale and
>> resample, then re-apply the color palette. The best resampling
>> algorithm is bicubic interpolation (see my chapter in the Pawley book,
>> where I compare algorithms). Note that when converting the image to
>> grayscale you must do it by applying a grayscale palette while
>> maintaining indices, not using 'closest match' which will give you
>> nonsense.
>>
>> Generally I don't use Photoshop, I use Paint Shop Pro, but
>> the basic steps should be similar. Generate a gray-scale image with all
>> 256 values present, and save the palette as a palette file (I call it
>> lin_grey). Open the user's image and save the palette as another
>> palette file. Now replace the palette, maintaining indices, by
>> lin_grey. You now have a grayscale image which you can scale
>> effectively. Scale with bicubic interpolation to the required value,
>> then re-apply the original palette (again, of course, maintaining
>> indices).
>>
>> Is this kosher? Absolutely, 100%. Your image IS a
>> gray-scale image, the palette is just an add-on. It's no different from
>> enlarging a negative in a darkroom enlarger. In the end your picture
>> will be represented in the printed page by an array of dots, between 120
>> and 300 to the inch, depending on the quality of the journal. Your goal
>> is to get the data to convert accurately to this representation.
>>
>> Guy
>>
>> Optical Imaging Techniques in Cell Biology
>> by Guy Cox CRC Press / Taylor & Francis
>> http://www.guycox.com/optical.htm
>> ______________________________________________
>> Associate Professor Guy Cox, MA, DPhil(Oxon)
>> Australian Centre for Microscopy & Microanalysis,
>> Madsen Building F09, University of Sydney, NSW 2006
>>
>> Phone +61 2 9351 3176 Fax +61 2 9351 7682
>> Mobile 0413 281 861
>> ______________________________________________
>> http://www.guycox.net
>>
>>
>> -----Original Message-----
>> From: Confocal Microscopy List [mailto:[log in to unmask]]
>> On Behalf Of David Knecht
>> Sent: Thursday, 8 September 2011 8:25 PM
>> To: [log in to unmask]
>> Subject: Re: Preparing figures for publication --PPI vs DPI
>>
>> *****
>> To join, leave or search the confocal microscopy listserv, go to:
>> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>> *****
>>
>> I have been forced back into this issue in helping someone get an image
>> taken on our microscopes through a journal editor. The images were
>> fairly low resolution. They were manipulated in Photoshop, imported
>> into Canvas, text added and then output for the journal as a figure at
>> 300dpi TIFF's in photoshop. What we noticed is that in Photoshop, when
>> the images were indexed color, the up sampling was done by taking each
>> pixel and subdividing it into smaller pixels. Thus the pixelation of
>> the original was maintained and the editor did not like that. What they
>> wanted done was to convert the images to RGB and then upsample. In that
>> case (or grayscale), Photoshop does an interpolation making a smoother
>> looking zoomed in view, but that is a change in the data. The index
>> version is actually more accurate although less pleasing. I plan to
>> argue that this is not necessary or desirable, but I did not know that
>> Photoshop (but not ImageJ) makes this distinction and thought others
>> might want to know it happens. Does anyone know why index vs. RGB
>> should matter to the up sampling algorithm. THanks- Dave
>>
>>
>> On Jul 13, 2010, at 4:59 PM, Armstrong, Brian wrote:
>>
>>> Well, as George McNamara has suggested many times on the list the
>> image you should send in the "mean image", (an example of the images
>> collected).
>>> I would predict that this is rarely the case and that most images sent
>> to journals are extreme examples of the best possible image collected
>> and corrected.
>>> However, sending the entire image set (in Gigabytes or even Terabytes)
>> to the journal seems impractical. Perhaps instead one could acquire the
>> entire data set by e-mailing the author.
>>>
>>>
>>> Brian D Armstrong PhD
>>> Light Microscopy Core Manager
>>> Beckman Research Institute
>>> City of Hope
>>> Dept of Neuroscience
>>> 1450 E Duarte Rd
>>> Duarte, CA 91010
>>> 626-256-4673 x62872
>>>
>> http://www.cityofhope.org/research/support/Light-Microscopy-Digital-Imag
>> ing/Pages/default.aspx
>>> From: Confocal Microscopy List
>> [mailto:[log in to unmask]] On Behalf Of Andreas
>> Bruckbauer
>>> Sent: Tuesday, July 13, 2010 2:14 AM
>>> To: [log in to unmask]
>>> Subject: Re: Preparing figures for publication --PPI vs DPI
>>>
>>> Dear all,
>>> so far this discussion mainly focusses on "the image" but one image
>> alone can not represent the data. Think about the data set represented
>> by a distribution which might be gaussian, so you would need at
>> least the mean and the width of the distribution. What i want to say
>> is that you will need to do a quantification and then show an image
>> representing the mean and show a histogram which gives the reader a
>> clue about how the data is distributed. There might be more than one
>> variable which is important for the analysis and of course you should
>> do an idependend repeat of the experiment.
>>> So what would you suggest, sending all the original image files to the
>> data bank?
>>> I still think that rather than re-analysing other scientists data a
>> repeat of the experiment in another lab is more important, there are a
>> lot of thinks apart from data analysis and representation which can go
>> wrong. However i see the point that when the paper is about image
>> analysis to provide the original files to let other groups repreat the
>> analysis.
>>>
>>> best wishes
>>>
>>> Andreas
>>>
>>>
>>>
>>>
>>> -----Original Message-----
>>> From: Daniel James White <[log in to unmask]>
>>> To: [log in to unmask]
>>> Sent: Tue, 13 Jul 2010 9:01
>>> Subject: Re: Preparing figures for publication --PPI vs DPI
>>>
>>> Dear Jerry,
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> cheers for adding your valuable input to this discussion.
>>>
>>>
>>>
>>>
>>> Its been a long and interesting one so far.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> On Jul 13, 2010, at 7:01 AM, CONFOCALMICROSCOPY automatic digest
>> system wrote:
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>> Date: Mon, 12 Jul 2010 12:47:51 -0500
>>>
>>>
>>>
>>>
>>>> From: "Jerry (Gerald) Sedgewick" <[log in to unmask]>
>>>
>>>
>>>
>>>
>>>> Subject: Re: Preparing figures for publication --PPI vs DPI
>>>
>>>
>>>
>>>
>>>>
>>>
>>>
>>>
>>>
>>>> <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
>>>
>>>
>>>
>>>
>>>> <html>
>>>
>>>
>>>
>>>
>>>> <head>
>>>
>>>
>>>
>>>
>>>> <meta content="text/html; charset=windows-1252"
>>>
>>>
>>>
>>>
>>>> http-equiv="Content-Type">
>>>
>>>
>>>
>>>
>>>> </head>
>>>
>>>
>>>
>>>
>>>> <body bgcolor="#ffffff" text="#000000">
>>>
>>>
>>>
>>>
>>>> Hi All,<br>
>>>
>>>
>>>
>>>
>>>> <br>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> maybe you might like to turn off HTML text formatting in your email
>> client when
>>>
>>>
>>>
>>>
>>> you post to the lists
>>>
>>>
>>>
>>>
>>> in this case its not so distracting but sometimes emails to this an
>> other list
>>>
>>>
>>>
>>>
>>> as unreadable in digest mode
>>>
>>>
>>>
>>>
>>> due to the thousands of html tags.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>> I'm a little late in addressing this issue, but the PPI/DPI part of
>>>
>>>
>>>
>>>
>>>> this conversation is misleading. What I mean to say is that the
>> real
>>>
>>>
>>>
>>>
>>>> issue may not be resolution, but reproduction.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> A very good point.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>> Images that are sent to
>>>
>>>
>>>
>>>
>>>> publishers with the full dynamic range of 0 - 255 pixel values may
>>>
>>>
>>>
>>>
>>>> likely be reproduced with no details at the bright end (240 - 255),
>> and
>>>
>>>
>>>
>>>
>>>> the dark end (0 - 20). Printing presses cannot reproduce detail
>> within
>>>
>>>
>>>
>>>
>>>> these tonal ranges because of limitations with dropping ink on paper
>>>
>>>
>>>
>>>
>>>> without A) having the drop not stick when the tonal values are
>> bright
>>>
>>>
>>>
>>>
>>>> and B) having the inks blend into each other through capillary
>> action
>>>
>>>
>>>
>>>
>>>> at the darkest values. This phenomenon gets worse when the paper is
>>>
>>>
>>>
>>>
>>>> lower quality (e.g., "Science").<br>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> So what we are saying is maybe summarised as follows:
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> Small thumbnail images in print can only every be just that - pointers
>> to go and
>>>
>>>
>>>
>>>
>>> look at the real digital image on some online source.
>>>
>>>
>>>
>>>
>>> No one can hope to reach quantitative conclusions from a small print
>> image.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> We can probably consign print images on paper to the heap of old
>> technologies,
>>>
>>>
>>>
>>>
>>> which are no longer relevant to the work we do.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> Images in PDF files "could" be useful, but not if they are lossy
>> compressed and
>>>
>>>
>>>
>>>
>>> missing meta data, which they usually are.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>> <br>
>>>
>>>
>>>
>>>
>>>> Color reproduction is generally worse because primary and secondary
>>>
>>>
>>>
>>>
>>>> colors are used to show experimental evidence. These colors are not
>>>
>>>
>>>
>>>
>>>> always within the range of printing presses, and so these tend to
>> print
>>>
>>>
>>>
>>>
>>>> blobs sans detail when color choices are not appropriate for
>> publishing
>>>
>>>
>>>
>>>
>>>> onto paper.<br>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> To make matters worse....
>>>
>>>
>>>
>>>
>>> fors most peoples eyes, blue is much fainter than red and green, with
>> green
>>>
>>>
>>>
>>>
>>> often being precieved as brightest.
>>>
>>>
>>>
>>>
>>> So forget even semi quantitative comparisons of intensity there.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> In any case we dont print in red green and blue, so these are silly
>> choices for
>>>
>>>
>>>
>>>
>>> a print version of figures.
>>>
>>>
>>>
>>>
>>> The print colour space would be better, Magenta, yellow and cyan....
>> but still
>>>
>>>
>>>
>>>
>>> for our quantitative purposed not good.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> Screen (LCD/CRT) colours are more similar to out eyes trichromatic
>> response.
>>>
>>>
>>>
>>>
>>> That why they are red green and blue... but the blue looks darker then
>> green
>>>
>>>
>>>
>>>
>>> problem still persists.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>> <br>
>>>
>>>
>>>
>>>
>>>> More often than not, the issue of whether or not details can be
>>>
>>>
>>>
>>>
>>>> resolved by eye on a printed page is not that of how many pixels
>> exist
>>>
>>>
>>>
>>>
>>>> in the submitted image, but in how effectively tones and colors were
>>>
>>>
>>>
>>>
>>>> fitted to the printing press. This is especially true if the pdf
>> image
>>>
>>>
>>>
>>>
>>>> at non-zoomed, computer screen resolution reveals desired details,
>> but
>>>
>>>
>>>
>>>
>>>> the printed page does not: it is more likely that a 133 line per
>> inch
>>>
>>>
>>>
>>>
>>>> screened image appearing in publication has more resolution than the
>>>
>>>
>>>
>>>
>>>> computer screen (often figured at an average of 90 pixels per inch,
>>>
>>>
>>>
>>>
>>>> with 72 pixels being the "old" standard). The color and tonal range
>> of
>>>
>>>
>>>
>>>
>>>> reproduction of a computer screen is greater than on a printing
>> press.<br>
>>>
>>>
>>>
>>>
>>>> <br>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> Indeed it is, and so is a better visualisation tool...
>>>
>>>
>>>
>>>
>>> but to get the most out of it, we also need to be smart.
>>>
>>>
>>>
>>>
>>> We cant compare the brightness of blue and green due to our
>> physiology,
>>>
>>>
>>>
>>>
>>> even if the screen is properly calibrated (usually not the case).
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> So, we should not ry to do that on screen or in print.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> Our eyes are much better at discriminating greyscale brightness scale,
>>
>>>
>>>
>>>
>>>
>>> but actually we are sill pretty bad at that.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> see this example of where we are easily fooled when trying to compare
>> grey scale
>>>
>>>
>>>
>>>
>>> brightness.
>>>
>>>
>>>
>>>
>>> http://web.mit.edu/persci/people/adelson/checkershadow_illusion.html
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> The situation with colouyr is more complicated, and also beset with
>> pitfalle
>>>
>>>
>>>
>>>
>>> due to the way our optical system works... it tries to find contrast
>> in any
>>>
>>>
>>>
>>>
>>> scene,
>>>
>>>
>>>
>>>
>>> and even generates contrast or colour differences that are not there
>> (but gave
>>>
>>>
>>>
>>>
>>> us an evolutionary advantage in picking fruit)
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> See the spiral image at the bottom f this colocalisation tutorial,
>>>
>>>
>>>
>>>
>>> which shows us that we should be careful when interpreting colour
>> merge multi
>>>
>>>
>>>
>>>
>>> channel images.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> http://pacific.mpi-cbg.de/wiki/index.php/Colocalization_Analysis
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> you can draw the spiral pattern yourself in imageJ with the marco
>> script there,
>>>
>>>
>>>
>>>
>>> and prove to yourself that your eyes lie to you.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> Greyscale images can be made interpretable in a semi quantitative
>> manner by
>>>
>>>
>>>
>>>
>>> using for instance the fire colour look up table.
>>>
>>>
>>>
>>>
>>> Each bright colour represents some intensity value, so its easier to
>> compare
>>>
>>>
>>>
>>>
>>> intensities over and between images.
>>>
>>>
>>>
>>>
>>> Physicists and chemists do this all the time, and think we are mad for
>> showing
>>>
>>>
>>>
>>>
>>> DAPI staining in black to Blue - because "thats what it looks like"
>>>
>>>
>>>
>>>
>>> They are right , we are mad to do that.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>> Like others before this email, I believe that it is best that
>>>
>>>
>>>
>>>
>>>> scientists take the task of reproduction as much in their own hands
>> as
>>>
>>>
>>>
>>>
>>>> possible so that the outcome can be controlled. The image is a
>>>
>>>
>>>
>>>
>>>> reproduction of what was once under a microscope, and it behooves
>> the
>>>
>>>
>>>
>>>
>>>> researcher to make that appear as close to the original
>> representation
>>>
>>>
>>>
>>>
>>>> as possible. <br>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> I'm not sure that this is ever going to be a helpful way to approach
>> the
>>>
>>>
>>>
>>>
>>> problem.
>>>
>>>
>>>
>>>
>>> No matter how carefully you set out your images and send them to the
>> publisher,
>>>
>>>
>>>
>>>
>>> you have no control and what they do with them next. None.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> The only way around this is to have the images in the paper act only
>> as
>>>
>>>
>>>
>>>
>>> thumbnails
>>>
>>>
>>>
>>>
>>> which point to the original image data file(s) on an online
>> repository,
>>>
>>>
>>>
>>>
>>> like the JCB image viewer or our wished for public biological image
>> database...
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> An image is not the sample, it always contain much less info that the
>> sample
>>>
>>>
>>>
>>>
>>> did.
>>>
>>>
>>>
>>>
>>> The trick is to keep the useful info as it passes through the scope
>> lenses,
>>>
>>>
>>>
>>>
>>> the detector, and the computer you your brain/eyes.
>>>
>>>
>>>
>>>
>>> I agree, an image is a representation of the info that made it through
>> the
>>>
>>>
>>>
>>>
>>> microscope to you.
>>>
>>>
>>>
>>>
>>> Trying to represent it "faithfully" and as it "truly looks" are both
>> missing
>>>
>>>
>>>
>>>
>>> the mark.
>>>
>>>
>>>
>>>
>>> The image is usually degraded by the blur of the Point spread function
>> / OTF
>>>
>>>
>>>
>>>
>>> and by various sources of noise.
>>>
>>>
>>>
>>>
>>> Thus, the image is an artifact in of itself. What we want to know
>> about is the
>>>
>>>
>>>
>>>
>>> sample.
>>>
>>>
>>>
>>>
>>> The images contains info from the sample in a degraded and incomplete
>> from.
>>>
>>>
>>>
>>>
>>> We have to work around that, and not pretend that the image fully
>> represents the
>>>
>>>
>>>
>>>
>>> sample.
>>>
>>>
>>>
>>>
>>> It does not.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>> <br>
>>>
>>>
>>>
>>>
>>>> As far as images being data points, this also is true, and these are
>> to
>>>
>>>
>>>
>>>
>>>> remain unaltered (unless flatfield correcting, background
>> subtracting,
>>>
>>>
>>>
>>>
>>>> etc) for measuring. A faithful representation of that image when
>>>
>>>
>>>
>>>
>>>> reproduced is another matter altogether.<br>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> Indeed it is, and i think there is even no such thing as a faithful
>>>
>>>
>>>
>>>
>>> representation of the Sample as an image,
>>>
>>>
>>>
>>>
>>> so the faithful representation of the image is also then something to
>> think
>>>
>>>
>>>
>>>
>>> about.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> I think what matters is How you represent the image data....
>>>
>>>
>>>
>>>
>>> in some cases an illustration like Hooke's drawing of cells might even
>> be better
>>>
>>>
>>>
>>>
>>> than a digital representation,
>>>
>>>
>>>
>>>
>>> if you want to get a certain message across to the reader. Original
>> digital
>>>
>>>
>>>
>>>
>>> image data available too of course online.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> cheers
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> Dan
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>> <br>
>>>
>>>
>>>
>>>
>>>> Cheers,<br>
>>>
>>>
>>>
>>>
>>>> <br>
>>>
>>>
>>>
>>>
>>>> Jerry Sedgewick<br>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> Dr. Daniel James White BSc. (Hons.) PhD
>>>
>>>
>>>
>>>
>>> Senior Microscopist / Image Visualisation, Processing and Analysis
>>>
>>>
>>>
>>>
>>> Light Microscopy and Image Processing Facilities
>>>
>>>
>>>
>>>
>>> Max Planck Institute of Molecular Cell Biology and Genetics
>>>
>>>
>>>
>>>
>>> Pfotenhauerstrasse 108
>>>
>>>
>>>
>>>
>>> 01307 DRESDEN
>>>
>>>
>>>
>>>
>>> Germany
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> +49 (0)15114966933 (German Mobile)
>>>
>>>
>>>
>>>
>>> +49 (0)351 210 2627 (Work phone at MPI-CBG)
>>>
>>>
>>>
>>>
>>> +49 (0)351 210 1078 (Fax MPI-CBG LMF)
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> http://www.bioimagexd.net BioImageXD
>>>
>>>
>>>
>>>
>>> http://pacific.mpi-cbg.de Fiji - is just ImageJ (Batteries
>> Included)
>>>
>>>
>>>
>>>
>>> http://www.chalkie.org.uk Dan's Homepages
>>>
>>>
>>>
>>>
>>> https://ifn.mpi-cbg.de Dresden Imaging Facility Network
>>>
>>>
>>>
>>>
>>> dan (at) chalkie.org.uk
>>>
>>>
>>>
>>>
>>> ( white (at) mpi-cbg.de )
>>>
>>>
>>>
>>>
>>>
>>> ---------------------------------------------------------------------
>>> SECURITY/CONFIDENTIALITY WARNING:
>>> This message and any attachments are intended solely for the
>> individual or entity to which they are addressed. This communication may
>> contain information that is privileged, confidential, or exempt from
>> disclosure under applicable law (e.g., personal health information,
>> research data, financial information). Because this e-mail has been sent
>> without encryption, individuals other than the intended recipient may be
>> able to view the information, forward it to others or tamper with the
>> information without the knowledge or consent of the sender. If you are
>> not the intended recipient, or the employee or person responsible for
>> delivering the message to the intended recipient, any dissemination,
>> distribution or copying of the communication is strictly prohibited. If
>> you received the communication in error, please notify the sender
>> immediately by replying to this message and deleting the message and any
>> accompanying files from your system. If, due to the security risks, you
>> do not wish to receive further communications via e-mail, please reply
>> to this message and inform the sender that you do not wish to receive
>> further e-mail from the sender.
>>> ---------------------------------------------------------------------
>>
>> Dr. David Knecht
>> Department of Molecular and Cell Biology
>> Co-head Flow Cytometry and Confocal Microscopy Facility
>> U-3125
>> 91 N. Eagleville Rd.
>> University of Connecticut
>> Storrs, CT 06269
>> 860-486-2200
>> 860-486-4331 (fax)
>
--
Dr. Dmitry Sokolov
Institute of Fundamental Sciences
Massey University, Palmerston North
New Zealand
|
