My quest to "fine tune" and understand my setup.
Moderators: rjlittlefield, ChrisR, Chris S., Pau
5. Use an extra aperture in the back of the Raynox.
In this test I use an extra aperture in the back of the Raynox lens. One with 18 mm aperture and the other with 10 mm aperture.
One problem with the extra aperture is quit a lot of vignetting.
Examples with zoom=24 mm, 50 mm, 107 mm and 215 mm
First without extra aperture:
With 18 mm extra aperture:
And finally 10 mm aperture:
And a comparison of 100% crops. All images are stacked.
From left to right: No aperture, 18 mm aperture, 10 mm aperture.
Zoom 24 mm
Zoom 50 mm
Zoom 107 mm
Zoom 215 mm
No imrovement with shorter zoom settings but a significant improvement at full zoom where the normal result is really bad.
Next step will be to check at what zoom settings it starts to improve the result and then it will be on to teleconverter 2x
/Leif K
In this test I use an extra aperture in the back of the Raynox lens. One with 18 mm aperture and the other with 10 mm aperture.
One problem with the extra aperture is quit a lot of vignetting.
Examples with zoom=24 mm, 50 mm, 107 mm and 215 mm
First without extra aperture:
With 18 mm extra aperture:
And finally 10 mm aperture:
And a comparison of 100% crops. All images are stacked.
From left to right: No aperture, 18 mm aperture, 10 mm aperture.
Zoom 24 mm
Zoom 50 mm
Zoom 107 mm
Zoom 215 mm
No imrovement with shorter zoom settings but a significant improvement at full zoom where the normal result is really bad.
Next step will be to check at what zoom settings it starts to improve the result and then it will be on to teleconverter 2x
/Leif K
6. Use cameras built-in teleconverter
The camera has a built-in digital Teleconverter (1,5x or 2x).
This test is to compare the best images from "test 5", with images of the same field of view but "zoomed out" combined with the 2x Teleconverter.
From left to right: Best image test 5, 2x TC, 2x TC + 10 mm extra aperture.
Zoom 24 mm without TC and 11 mm with TC
Zoom 50 mm without TC and 29 mm with TC
Zoom 107 mm without TC and 50 mm with TC
Zoom 215 mm without TC and 80 mm with TC
And finally at max zoom (215 mm) + 2x TC.
Right one with 10 mm extra aperture.
I set the zoom "by hand" and didnt hit the exakt setting to match 2x but my conclusion is to avoid the built-in Teleconverter unless I need a higher magnification than full zoom or if I want to do a quicker stack with less focus steps.
The camera has a built-in digital Teleconverter (1,5x or 2x).
This test is to compare the best images from "test 5", with images of the same field of view but "zoomed out" combined with the 2x Teleconverter.
From left to right: Best image test 5, 2x TC, 2x TC + 10 mm extra aperture.
Zoom 24 mm without TC and 11 mm with TC
Zoom 50 mm without TC and 29 mm with TC
Zoom 107 mm without TC and 50 mm with TC
Zoom 215 mm without TC and 80 mm with TC
And finally at max zoom (215 mm) + 2x TC.
Right one with 10 mm extra aperture.
I set the zoom "by hand" and didnt hit the exakt setting to match 2x but my conclusion is to avoid the built-in Teleconverter unless I need a higher magnification than full zoom or if I want to do a quicker stack with less focus steps.
Last edited by skrylten on Sat Sep 03, 2016 3:45 am, edited 1 time in total.
7. Flash setting and shutterspeed
The camera has 3 different flash powers, low, medium and full.
Its possible to use first or second curtain sync. Its an electronic shutter and I think the curtains specify how to read the pixels from the sensor.
I´m using second curtain but I dont think it makes any difference in my set up.
With my milk-bottle diffuser the low flash power doesnt give enough light so I have to stick with medium and full power.
Slower shutter speeds ( longer than ~ 1/1000 s) doesnt have any impact on the image (its all about the flash) but I can use CHDK to override the shutterspeed when I need faster shutter speeds.
Here are some examples of single images with correpsonding histograms.
F 7,1, ISO 80, Zoom 129 mm
From left to right:
- Medium flash 1/1000 s
- Medium flash 1/40000 s
- Full flash 1/6400 s
- Full flash 1/10000 s
- Full flash 1/40000 s
Which one to choose and why ? Maybe it doesnt matter ?
This is 100 % crops with adjusted brightness and contrast and with corresponding histograms.
Looking forward to input on how to choose
The camera has 3 different flash powers, low, medium and full.
Its possible to use first or second curtain sync. Its an electronic shutter and I think the curtains specify how to read the pixels from the sensor.
I´m using second curtain but I dont think it makes any difference in my set up.
With my milk-bottle diffuser the low flash power doesnt give enough light so I have to stick with medium and full power.
Slower shutter speeds ( longer than ~ 1/1000 s) doesnt have any impact on the image (its all about the flash) but I can use CHDK to override the shutterspeed when I need faster shutter speeds.
Here are some examples of single images with correpsonding histograms.
F 7,1, ISO 80, Zoom 129 mm
From left to right:
- Medium flash 1/1000 s
- Medium flash 1/40000 s
- Full flash 1/6400 s
- Full flash 1/10000 s
- Full flash 1/40000 s
Which one to choose and why ? Maybe it doesnt matter ?
This is 100 % crops with adjusted brightness and contrast and with corresponding histograms.
Looking forward to input on how to choose
- rjlittlefield
- Site Admin
- Posts: 23605
- Joined: Tue Aug 01, 2006 8:34 am
- Location: Richland, Washington State, USA
- Contact:
Right. The main difference would be if you had a moving subject plus enough ambient light to register during the exposure. In that case second-curtain sync will flash the subject at the end of the exposure, so any streak from the ambient illumination will appear "behind" the main exposure in terms of movement. This is the effect that we normally want to see. In contrast, first-curtain sync will flash the subject at the start of the exposure, so any streak from the ambient exposure will be captured later and will appear "in front of" the main exposure in terms of movement. This looks odd. See the illustrations at http://neilvn.com/tangents/first-curtai ... tain-sync/ .skrylten wrote:Its possible to use first or second curtain sync. Its an electronic shutter and I think the curtains specify how to read the pixels from the sensor.
I´m using second curtain but I dont think it makes any difference in my set up.
--Rik
8. How to align images for noise reduction and superresolution
This post should be about aligning images but I will start by having a look at a single image and try to get rid om some abberations (if possible).
The test images are of som printed dots on white paper.
Zoom 164 mm, F 6.3 and 1/8000 s
First the whole image followed by 100% crops of a dot in the center and a dot in the corner.
A lot of transverse chromatic abberation in the corners ...
Another intresting comparison is to split the image into the different color channels.
From left to right the: red, green and blue channel of 4 consecutive focus steps.
To me it looks like there is a full focus step of distance between the different color channels in axial chromatic aberration.
My first thought was to combine color channels from different focus steps to combine a new improved image but I think it should be even better to stack each channel separately and then combine them to a new image.
Here are 100% crops of a single image followed by a "normal" stack and finally a stack with color channels stacked separately.
And here is the difference of a dot in the corner. The red "top" is from aligning the color channels.
It seems like its possible to improve on both axial and transverse chromatic abberation with this aproach and get a stacked image that "looks better" than what can be seen in the focused part of a single image. That was interesting for me ...
This post should be about aligning images but I will start by having a look at a single image and try to get rid om some abberations (if possible).
The test images are of som printed dots on white paper.
Zoom 164 mm, F 6.3 and 1/8000 s
First the whole image followed by 100% crops of a dot in the center and a dot in the corner.
A lot of transverse chromatic abberation in the corners ...
Another intresting comparison is to split the image into the different color channels.
From left to right the: red, green and blue channel of 4 consecutive focus steps.
To me it looks like there is a full focus step of distance between the different color channels in axial chromatic aberration.
My first thought was to combine color channels from different focus steps to combine a new improved image but I think it should be even better to stack each channel separately and then combine them to a new image.
Here are 100% crops of a single image followed by a "normal" stack and finally a stack with color channels stacked separately.
And here is the difference of a dot in the corner. The red "top" is from aligning the color channels.
It seems like its possible to improve on both axial and transverse chromatic abberation with this aproach and get a stacked image that "looks better" than what can be seen in the focused part of a single image. That was interesting for me ...
8. How to align images for noise reduction and superresolution, part 2
The small sensor in the SX50 is quite noisy but I can reduce the noise by taking several images at each focus step and use the median value of them to create a noise reduced image.
This test is to see if the images need to be aligned and if a "simple" alignment is enough.
The test image is the green colorchannel of a butterfly wing.
Zoom 107 mm, F 6,3, 1/8000 s, 100 % crops
First a single image.
And next noise reduced images with different alignment of 9 separate images.
From left to right:
1. No alignment.
2. Standard alignment with align_image_stack ( downscaled by factor 2, 5x5 grid with 8 control points per grid.
3. align_image_stack with no downscaling, 10x10 grid with 16 control points per grid.
I think there is need for alignment but the standard alignment will be enough.
I need to type a lot of commands and parameters with these tests and to make my life easier I have made an "embryo" of a "Stacking GUI" based on a hta file and VBscripts. I will try to update this GUI along with the tests .
The small sensor in the SX50 is quite noisy but I can reduce the noise by taking several images at each focus step and use the median value of them to create a noise reduced image.
This test is to see if the images need to be aligned and if a "simple" alignment is enough.
The test image is the green colorchannel of a butterfly wing.
Zoom 107 mm, F 6,3, 1/8000 s, 100 % crops
First a single image.
And next noise reduced images with different alignment of 9 separate images.
From left to right:
1. No alignment.
2. Standard alignment with align_image_stack ( downscaled by factor 2, 5x5 grid with 8 control points per grid.
3. align_image_stack with no downscaling, 10x10 grid with 16 control points per grid.
I think there is need for alignment but the standard alignment will be enough.
I need to type a lot of commands and parameters with these tests and to make my life easier I have made an "embryo" of a "Stacking GUI" based on a hta file and VBscripts. I will try to update this GUI along with the tests .
9. How to handle "inversed perspective", part 1
This is the post that so far has given me most headache to wrap my head around. Part 1 will be to describe the problem and have a first shot to handle the strange perspective.
First some images to describe the problem. 3 images at each zoom setting:
Left: focused at closest possible focus point.
Middle: Focused "midway to infinity".
Right: Focused at infinity.
Zoom 52 mm
Zoom 103 mm
Zoom 193 mm
At full zoom the subject is 50% bigger at infinity focus than at closest focus point.
To se the effect on a stacked image I did a stack with 103 images at full zoom (215 mm) of an Iphone 4 screen.
In the stack it seems like the screen is curved and closest to the camera at the right side
My idea to improve the result is to scale the individual images based on zoom setting and focus.
In this first attempt I used a third order equation to scale the images and got this result (crop to the right)
Better but very flat compared to the angle of the Iphone4 screen.
Next test was to add more scaling in an attempt to get some perspective in the image.
Now it looks better but its obvious the third order equation isn´t correct.
In part 2 I will try to improve the scaling of the images to get a more linear result in the stack.
/Leif K
This is the post that so far has given me most headache to wrap my head around. Part 1 will be to describe the problem and have a first shot to handle the strange perspective.
First some images to describe the problem. 3 images at each zoom setting:
Left: focused at closest possible focus point.
Middle: Focused "midway to infinity".
Right: Focused at infinity.
Zoom 52 mm
Zoom 103 mm
Zoom 193 mm
At full zoom the subject is 50% bigger at infinity focus than at closest focus point.
To se the effect on a stacked image I did a stack with 103 images at full zoom (215 mm) of an Iphone 4 screen.
In the stack it seems like the screen is curved and closest to the camera at the right side
My idea to improve the result is to scale the individual images based on zoom setting and focus.
In this first attempt I used a third order equation to scale the images and got this result (crop to the right)
Better but very flat compared to the angle of the Iphone4 screen.
Next test was to add more scaling in an attempt to get some perspective in the image.
Now it looks better but its obvious the third order equation isn´t correct.
In part 2 I will try to improve the scaling of the images to get a more linear result in the stack.
/Leif K
- rjlittlefield
- Site Admin
- Posts: 23605
- Joined: Tue Aug 01, 2006 8:34 am
- Location: Richland, Washington State, USA
- Contact:
Not sure if this will help solve your problem, but see http://www.photomacrography.net/forum/v ... 0750#50750 for an explanation of why your perspective is inverted.
--Rik
--Rik
9. How to handle "inversed perspective", part 2
Thanks for the link to that post, Rik. It really gives me a better perspective on my issue
By scaling the images I am now able to get a flat view of the stack or some perspective that maybe makes it easier to understand the geometry of the subject.
Left: Stack without scaling of images.
Middle: Scaled to get a flat view
Right: Scaled to get a perspective.
Thanks for the link to that post, Rik. It really gives me a better perspective on my issue
By scaling the images I am now able to get a flat view of the stack or some perspective that maybe makes it easier to understand the geometry of the subject.
Left: Stack without scaling of images.
Middle: Scaled to get a flat view
Right: Scaled to get a perspective.
10. Noise reduction with several images/step
There is quite a lot of noise in a single image but by taking several images at each step and combine them into one image I can reduce the noise.
Here is a comparison of a single image and the median value of 3, 5, 9, 15 and 27 images. They are close to 6x their actual size.
Images taken at zoom 83 mm.
Its possible to see an improvement at least up to 27 images per step.
There is quite a lot of noise in a single image but by taking several images at each step and combine them into one image I can reduce the noise.
Here is a comparison of a single image and the median value of 3, 5, 9, 15 and 27 images. They are close to 6x their actual size.
Images taken at zoom 83 mm.
Its possible to see an improvement at least up to 27 images per step.