[Unofficial] New Module: Sharpen

[Carlos Milovic]

Hmmm... I though that I fixed the 8 ways filters I was receiving a memory violation when a non integer radius was set. Was this the case? (maybe I uploaded an earlier compilation).


Thanks Carlos for your words


Harry, if you have any suggestion on simple processes that are worth implementing, just say so, and I'll add it to my To-do list. I like these simple tools, because I'm a bit rusty and need some previous "exercises" to get into my "programing shape"


 
BTW, I don't have a W94 computer... so I cannot help there... I'm counting on Juan's aid here

[Carlos Milovic]

Ok. I got the same message now, with any radius. Hmmmmmmmmmmm. Back to the debugger table

[Nocturnal]

Hi,

I compiled the Sharpen module for x64:

http://www.tungstentech.com/files/Sharpen-pxm-win64.zip

[Harry page]

Hi Sander

Many thanks for your skilled work, but I am afraid its still bugged

Regards Harry

[Nocturnal]

Luckily I did not attempt to fix any bugs. I just recompiled

[Harry page]

Hi Carlos

I would like a heal tool !  great for zapping hot pixels, I know there is a clone tool , but a heal tool ( Sorry photoshop talk ) gives a seamless fix  

And how about a dust bunny eradicator  ( This might be hard )

Also there was a hint but real time view of the sharpen tool

And ...........   no I will stop there as I will drive you mad


Harry

[Carlos Milovic]

I gave the code a quick look, but did not find that little $!"·$ bug I fear that I'll have a "nice" day ahead tomorrow, polishing my hunting skills


Harry, I'm thinking about that heal tool. I guess that it should not be that hard, working over the clone stamp code (although it is quite complex for me). Perhaps a simple statistical analysis may do the work, like I do in ABE's inner rejection algorithm. Another option is just a simple threshold. My problem is that either way the stars have small chances to survive, so I want to sophisticate that a bit.
What is a dust bunny? The artifact created by dust over the sensor? Sounds interesting... Maybe a FFT approach...
About RT, this kind of objects fly more than a pair of meters above my head I'll need Juan's help there.

Well, let's see what happens this weekend...

[Carlos Milovic]

I think I cornered him... at least, how to reproduce the error, and where it is NOT on the code The error appear only when working with floating points images, with the Laplace8 and WeightedLaplace8 algorithm. The answer seems straightforward then, but all algorithms share the same way of handling image bit depth, what is done at the beggining. So, I'm not sure of the reason of this strange behavior. I noticed no problems at all with integer samples...
Also there is something else that is annoying me. If there is no error message, the output is not always the same. I got some "hot" pixels (or even lines), randomly.

Juan, could you give me a hand here?

[Juan Conejero]

Hi Carlos,

Sure thing. However I ask for a bit of patience since right now I am involved in a huge mess - I am switching from VMware to VirtualBox, and this means I have to rebuild all Windows virtual machines. PixInsight's build system is "under construction" right now on all platforms 8) (non-developer users: the situation is under control -- reasonably )

[Juan Conejero]

Carlos, one more thing. Could you please upload your source code as a .zip archive? Currently tar is something too exotic for my Linux box 

[Carlos Milovic]

Ok. Let me finish some changes I'm doing to GaussianBlur (more parameters and a control window drawing the filter), and I'll upload both modules/codes. Sharpen has some minor changes too, mainly small optimizations of the code.

[Harry page]

Hi Carlos

Looks like you are having fun   

You are right a dust Bunnie is the shadow from a bit of muck,

I think there are two way of looking at this ,  try and model the mote and correct it, this would be very advanced and perhaps to much to ask or just look the motes in background areas and fix them with a fancy heal tool
Of course you may have many better ideas

Good luck with the hunt 

Harry

[Carlos Milovic]

Hi Harry

Yes, I'm thinking on a way to modelate them, automatically. Since it is a multiplicative effect, and with a somehow predictable shape (well, we know that it is very simetrical, and that is a "negative" effect. I've got two candidates to explore: fourier transform and homomorphic transform. Don't expect a quick answer, anyway First I want to finish/update other older modules I wrote but were not released.

[Carlos Milovic]

Juan (and anybody who wants to join the safari) , I just uploaded the last source code, compressed with zip format:

http://pteam.pixinsight.com/carlos/modules/Sharpen-src-2009-09-18.zip

[Carlos Milovic]

Hi all

I've updated the links above with the last release. Thanks to Juan's suggestions, I made it much more faster and robust, and it seems that the annoying bug is gone.