My quest to "fine tune" and understand my setup.

Just bought that first macro lens? Post here to get helpful feedback and answers to any questions you might have.

Moderators: rjlittlefield, ChrisR, Chris S., Pau

Grahame
Posts: 60
Joined: Tue Sep 30, 2014 11:36 pm
Location: New Zealand

Post by Grahame »

Hi Leif,
Have you ever tried a bit of software called PhotoAcute
http://photoacute.com/studio/
It combines multiple similar images to produce an image with higher res and less noise than a single image.
I had a play with the trial of the new version last week and it was surprising how well it worked.
It worked far better on raw images than jpegs, guess there's no surprise there.
I have thoughts of trying it for stacking, 10 images at each step or similar.
It does take a while to create each image so putting it in batch mode and going out for a while will be necessary.
It also has a few other tricks up it's sleeve.
Worth taking the trial for a spin.
Grahame

skrylten
Posts: 60
Joined: Fri Jan 03, 2014 1:41 pm

Post by skrylten »

Hi Grahame !

Thanks for the link. I haven´t tried it but it seems to do a lot of things. Some of them looks similar to the noise reduction and super resolution I´m trying out.

/Leif K

skrylten
Posts: 60
Joined: Fri Jan 03, 2014 1:41 pm

Post by skrylten »

11. Superresolution

The way I understand "super resolution" it would be possible to improve the resolution by taken several images, upsize them and combine them into a new image.

To test this with my setup I took 100 images of a butterfly wing. The camera is resting on the table so I presume their is some (at least) sub pixel moment between the images.

I didn´t want to use and process the complete image so I made a 400x300 crop of a part with some interesting scales.

Zoom 146 mm. Full image to the left and to the right the 400x300 crop.
ImageImage

Next step is to upsize the crops to 4000x3000 pixels and then combine them to one image with the median value of each pixel.
This is a comparison of the original crop and the median value of the upsized images.

Left: Original crop
Middle: "Median image" of upscaled images.
Right: "Median image" of upsized imaged (with imagemagick -resize)
Image

Thats some improvement but it turns out I can get the same result by using the median of the 100 individual crops and the resize the result to 4000x3000.
Image

My interpretation of this is that the improvement is only made up by noise reduction and not by "super resolution".
At least with my set-up I cant see any benefit from using lots of images combined with upscaling.

Maybe it could be useful in a set-up where it is possible to control subpixel movements and the camera/lens is resolving subjects down to pixel size ?

rjlittlefield
Site Admin
Posts: 23561
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

skrylten wrote:Maybe it could be useful in a set-up where it is possible to control subpixel movements and the camera/lens is resolving subjects down to pixel size ?
Yes. The key thing is that for super-resolution to work, the underlying optical image has to be higher resolution than what is captured by the sensor. In your case it appears that the optical image is blurred enough that the sensor is already capturing all the detail that's present.

--Rik

skrylten
Posts: 60
Joined: Fri Jan 03, 2014 1:41 pm

Post by skrylten »

12. Align images in the stack.

To align the images in the stack I use:

align_image_stack with no downscaling, 5x5 grid with 8 control points per grid.

The normal way is to align the images from the front to the back but I have seen that it can give some odd results if there are very few parts in focus in the first images.

To avoid this problem I'am doing two alignments, one from the middle image to the front of the stack and another one from the middle image to the back of the stack.

skrylten
Posts: 60
Joined: Fri Jan 03, 2014 1:41 pm

Post by skrylten »

13. Use slabs
14. Impact of diffent step sizes.
15. Stacking parameters
16. "Stacking tactics"

Stacking Part 1

A "normal" stack with enfuse gives a pretty decent result.
I mainly do stacks of flies and I can see some problems with transparency and halos with the hairs and bristles.

This is 100% crops of the stack of the "green images" to give some examples.
ImageImageImage

The hairs and brsitles are quite important when it comes to ID the fly so I wanted to decrease the transparency in the stacked result.
I decided to go for an approach where I use masks to find out the area in focus in each image. Its time consuming on my old laptop (I3 processor and 4G RAM) and takes around 6 min per image to get the mask.
Image

The result of this approach can be seen in the following 100% crops of the stack.
ImageImageImage

Its a better result and I think there is still room for some imrovement when it comes to preparing the masks.

Post Reply Previous topicNext topic