Re: Microscope Image Focus Quality plugin

[Krs5]

Just in case other people are interested in use of this plugin I have received the email below from Samuel Yang (see below) and the plugin works now as advertised when installed according to https://imagej.net/Microscope_Focus_Quality.

However, be aware that the plugin fails if the Stardist plugin is also installed. I guess there is a problem with duplicated file names but have not had a look at this.

Best wishes

Kees

________________________________
From: Samuel Yang <[hidden email]>
Sent: 14 May 2020 19:03
To: [hidden email] <[hidden email]>; Straatman, Kees (Dr.) <[hidden email]>
Subject: FIJI microscope image focus quality plugin fixed

Hi,

Curtis sent me your discussion here:
Bad URL Removed - see why - https://ees.sps.nih.gov/services/Pages/Anti-Virus.aspx%3FA2%3Dind2005%26L%3DIMAGEJ%26P%3D12850%26X%3DD6A84F280368AE06DC%26Y%3Dsamuely%2540google.com&data=02%7C01%7Ckrs5%40leicester.ac.uk%7Cc362d9121aa248c4ace708d7f83127c2%7Caebecd6a31d44b0195ce8274afe853d9%7C0%7C1%7C637250762322489402&sdata=rFASwELIUsEo7VwIQ4Pfl53FCpkT%2BY0ghmfSbcDnErI%3D&reserved=0>

Thanks for letting us know the model download link was broken, it should be fixed now.

Just a quick comment -- if you're finding the best-focus image in a z-stack, looking at simple statistics like the standard deviation of pixels in a slice is a robust and fast way.

But assessing image focus on a single image in isolation (without z-stacks, or without a large example dataset of in and out-of-focus images) is a harder task that we approached with our model:

https://ai.googleblog.com/2018/03/using-deep-learning-to-facilitate.html<https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fai.googleblog.com%2F2018%2F03%2Fusing-deep-learning-to-facilitate.html&data=02%7C01%7Ckrs5%40leicester.ac.uk%7Cc362d9121aa248c4ace708d7f83127c2%7Caebecd6a31d44b0195ce8274afe853d9%7C0%7C1%7C637250762322499397&sdata=RCvdTYT8qT%2FUFOdx%2FyGEkZIF%2BBfYjAiraxyBDg7meqo%3D&reserved=0>

PDF of paper:
http://rdcu.be/I5cE<https://eur03.safelinks.protection.outlook.com/?url=http%3A%2F%2Frdcu.be%2FI5cE&data=02%7C01%7Ckrs5%40leicester.ac.uk%7Cc362d9121aa248c4ace708d7f83127c2%7Caebecd6a31d44b0195ce8274afe853d9%7C0%7C1%7C637250762322499397&sdata=uxEwBLK9GusbNGIRPRr1yteI%2BrpSxIp0JlGfzfX5R%2Fo%3D&reserved=0>

It is slow, but you can choose to compute the measure on fewer tiles within an image:
[image.png]

Hope this helps. Feel free to share this response on the mailing list, I'm not sure how to post to it.

--
Samuel Yang


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