This is a truly amazing program. At a recent conference held by my firm, I sat in the audience, and I took a large number of indoor portraits at 200mm and ISO 1600 with a Nikon D70S. Using Neatimage, I was able to produce a large set of beautiful, virtually grain-free, 11 by 14s.
Now I'm trying to tackle a different problem. I take a large number of photos from behind a fence. I've attached a link with a typical image. Each time, there is an extremely regular pattern that looks like "graph paper" across the entire image. Is there a way to get neatimage to remove this?
Thanks for your help!
http://forrestmilder.smugmug.com/galler ... 3099/Large
Removing a regular pattern from photos
Removing a regular pattern from photos
-- Forrest Milder
Thank you for your kind words, I am glad NI helped you reduce noise in those images.
I am now looking at the sample image with the fence pattern and thinking that this is not as easy for NI as regular noise or grain, simply because this pattern is not what NI is expecting.
I think a specialized tool could be developed to eliminate fixed patterns like those often seen in scans of old photos. However in your images, the pattern is not exactly fixed so such a tool could have certain difficulties as well.
You could try to play with this pattern in Photoshop trying to produce/extract a mask that would cover all shadows produced by the fence and then apply some adjustments to the masked area. I am not sure this would completely eliminate the pattern but at least it could be made less noticeable.
Hope this helps.
Vlad
I am now looking at the sample image with the fence pattern and thinking that this is not as easy for NI as regular noise or grain, simply because this pattern is not what NI is expecting.
I think a specialized tool could be developed to eliminate fixed patterns like those often seen in scans of old photos. However in your images, the pattern is not exactly fixed so such a tool could have certain difficulties as well.
You could try to play with this pattern in Photoshop trying to produce/extract a mask that would cover all shadows produced by the fence and then apply some adjustments to the masked area. I am not sure this would completely eliminate the pattern but at least it could be made less noticeable.
Hope this helps.
Vlad
Removal of periodic elements
Forest - You may want to look into using a Fast Fourier Transform technique to suppress/eliminate a periodically occuring element within your images. This is not the easiest technique to appreciate intuitively, but it can be a fairly powerful way to isolate and manipulate a regularly occuring element within an image.
Please see:
http://www.nist.gov/lispix/imlab/labs.html
and scroll down to the FFT section for an example. NIH Image is a free image processing tool that is incredibly robust and, due to its target audience of scientists, a little confusing. It used to be available for us Mac users only, but now it appears that there is a PC version too - there is also a Java version called ImageJ. Do a Google search on FFT image processing to start exploring this technique - it may work for you. As with all image processing techniques, there is a trade-off - if you suppress the frequency of the mesh netting, you will also affect the visual content within the image of everything else at that frequency.
And, on a separate note, Neat Image rocks. I just learned of it yesterday, bought the Pro plug-in for Photoshop /Mac and my RAW workflow just got a whole lot better. Thanks NI Team.
Kirk
Please see:
http://www.nist.gov/lispix/imlab/labs.html
and scroll down to the FFT section for an example. NIH Image is a free image processing tool that is incredibly robust and, due to its target audience of scientists, a little confusing. It used to be available for us Mac users only, but now it appears that there is a PC version too - there is also a Java version called ImageJ. Do a Google search on FFT image processing to start exploring this technique - it may work for you. As with all image processing techniques, there is a trade-off - if you suppress the frequency of the mesh netting, you will also affect the visual content within the image of everything else at that frequency.
And, on a separate note, Neat Image rocks. I just learned of it yesterday, bought the Pro plug-in for Photoshop /Mac and my RAW workflow just got a whole lot better. Thanks NI Team.
Kirk
Example of Processing
Forest - I'm a smugmug user too and I have posted a quick attempt to solve your issue:
http://kirkt.smugmug.com/gallery/971871
with some more effort and attention to the FFT editing, you can probably do even better. As you can see, I took two steps:
1) perform the FFT of the original image, and edit the FFT to eliminate the unwanted frequency content - in this case, painting with black on the FFT is like erasing that frequency content - in the FFT, white represents high power of content at that frequency, black denotes no power at that frequency. The center of the FFT, the origin of the X and Y axes, represents the low frequency content - moving away from the center represents increasing frequency content. The "frequency we are talking about is spatial frequency.
You can see in the first comparison image that the quick painting of the FFT got rid of some of the mesh netting. Once paint black onto the FFT, you just perform the IFFT (inverse) and get the filtered image.
2) I then brought the image into photoshop and invoked the Neat Image filter and profiled the area highlighted in the second comparison image. It includes any noise, dirt texture and the intersections of the netting not eliminated in Step 1. It makes the dirt look a little too smooth, so i imagine you could play with the NI filter settings to counter this.
Hope this helps.
kirk
(Wow, that image processing class I took in grad school actually paid off!).
http://kirkt.smugmug.com/gallery/971871
with some more effort and attention to the FFT editing, you can probably do even better. As you can see, I took two steps:
1) perform the FFT of the original image, and edit the FFT to eliminate the unwanted frequency content - in this case, painting with black on the FFT is like erasing that frequency content - in the FFT, white represents high power of content at that frequency, black denotes no power at that frequency. The center of the FFT, the origin of the X and Y axes, represents the low frequency content - moving away from the center represents increasing frequency content. The "frequency we are talking about is spatial frequency.
You can see in the first comparison image that the quick painting of the FFT got rid of some of the mesh netting. Once paint black onto the FFT, you just perform the IFFT (inverse) and get the filtered image.
2) I then brought the image into photoshop and invoked the Neat Image filter and profiled the area highlighted in the second comparison image. It includes any noise, dirt texture and the intersections of the netting not eliminated in Step 1. It makes the dirt look a little too smooth, so i imagine you could play with the NI filter settings to counter this.
Hope this helps.
kirk
(Wow, that image processing class I took in grad school actually paid off!).
Thanks for your sugestions and illustration, Kirk. Sorry that it took so long for me to realize that you had posted something here!
Can you give me any more insight into just how hard it was to do this? I know how to use Neat Image, but not FFT. I'm not sure that I followed the description in your e-mail.
Thanks for any guidance you can offer.
-- Forrest
fmilder@nixonpeabody.com
Can you give me any more insight into just how hard it was to do this? I know how to use Neat Image, but not FFT. I'm not sure that I followed the description in your e-mail.
Thanks for any guidance you can offer.
-- Forrest
fmilder@nixonpeabody.com