Olympus auto distortion correction for 8mm lens

ArizonaMike

Mu-43 Top Veteran
Joined
Jan 19, 2018
Messages
817
Location
Arizona, USA
A couple of months ago I decided to try to use a fish-eye lens to replace the pano stitching I had been doing, and bought a Samyang 7.5mm. When the lens arrived I did some testing to see how well the de-fishing algorithm in Dxo's PhotoLab worked if I labeled the Samyang as an Olympus 8mm lens. It seemed to be "OK" as long as I could keep the subject level with the camera and keep the camera horizontal, but I had to add about 20% or more to the image height to keep the proportions right. I took the Samyang, along with my other lenses, on a recent trip to Arches National Park and Natural Bridges National Monument in Utah I was really disappointed. If the target of my camera shot was not level there was just an enormous amount of vertical compression.

I decided to try the Olympus 8mm and see if the new firmware for the E-M1.2 really would make that lens usable for pano and UWA shots, but this time I would do some testing to see how the lens, and the auto de-fishing in the EVF and the output jpgs, worked. The lens came in the other day and I have been busy testing and adjusting and wanted to share some of the information. I have not seen any posting of actual comparisons of the 8mm FE with the new firmware against other lenses and thought it might be helpful to others who are thinking of getting the lens.

The new firmware offers users the ability to turn the auto de-fish on and off, and offers the choice of 3 settings, 1, 2 or 3, with 1 being the widest and 3 being the narrowest. I was interested in several comparisons: (1) How does the Dxo PhotoLab distortion correction compare against Lightroom (6 in my case, not CC), (2) How do the Olympus auto de-fished jpgs compare against the Olympus 9-18 images, where the comparison is to see how the horizontal to vertical proportions match, and (3) How do the Olympus auto de-fished jpgs compare against the raw images when run through PL and LR. I also wanted some estimate as to what FOV the 3 settings actually gave me.

I apologize if this post is a bit long, but there are images to show and adjustment information to explain.

(1) PhotoLab vs Lightroom distortion correction.

This turned out to be simple. I ran the 8mm raw fisheye images through both products and loaded the jpgs into my pixel editor. I then copied one image on top of the other, reduced the opacity and aligned them. The two images were virtually identical, so there is no real difference between how the two applications correct the image distortion. However Lightroom insisted on keeping the original image sides ratio so I ended up losing all of the information in the output jpg beyond what could be displayed in a 4 x 3 image, and that was a lot. Perhaps there is some way to turn that restriction off in LR, but I could not find it.

PhotoLab-vs-Lightroom.jpg
Subscribe to see EXIF info for this image (if available)


As you can see, a lot of information was lost, and this was only the medium-wide setting.

(2) Olympus 8mm wide vs Olympus 9-18mm jpg

I used the Olympus 9-18mm lens as the basis to test how well the corrected fisheye images kept the proper horizontal to vertical proportions. With my Samyang the vertical parts of the image were noticeably compressed, and I wanted to see if that was also true with the Olympus 8mm.

I loaded both the 9-18mm and 8mm images into the editor, superimposed the narrower image on top of the wider one, lowered the opacity and adjusted the height and width of the 9-18 image to try to get a match with the 8mm image. I wanted to see how close the adjustments were to each other. If they were significantly different it meant that there was a significant amount of visual compression that was not being taken care of

Compare-9-18jpgToOlympusWide.jpg
Subscribe to see EXIF info for this image (if available)


The horizontal and vertical adjustments to the 9-18mm image were 66.6 and 67.8% respectively. That is, I needed to apply more horizontal reduction than vertical, which implies that the Olympus de-fishing algorithm also compresses the vertical part of the image more than the horizontal in its widest 8mm setting, but not much more compared to the Samyang. With that lens I had to adjust the height of the image 20% and more to compensate for the vertical compression.

It should be mentioned that the results depend upon the 9-18mm image being a good representation of the actual subject proportions.

(3) Olympus 8mm wide vs Olympus 9-18mm raw using PhotoLab

I basically did the same thing in this step as I did in step 2, although here i was comparing the Olympus 8mm auto de-fished image against the Olympus 9-18 image without having PhotoLab process the raw, and thus apply its distortion correction to the 9-18 image. I wanted to see how the auto de-fished image looked against the native 9-18 jpg. the results were very similar to the above. Here is the comparison:

Comparison-PL9-18vsPL8Raw.jpg
Subscribe to see EXIF info for this image (if available)


Here the adjustment figures for horizontal and vertical are 66.2 and 67.5%, again a very small difference. Considering that matching overlaying images depends upon subjective ideas about what matches, these figures are identical to the previous ones for all intents and purposes.

This testing was very limited, and not at all rigorous. The camera was mounted on a tripod, the settings for the different lenses and for the settings of the 8mm lens were almost exactly the same, but no laboratory accurate conditions existed and everything was done with the idea that the results would only give an indication as to what to expect. However some things were clear. Here is an in-camera adjusted jpg, shot with the middle adjustment:

8mm-Setting2-OOC.jpg
Subscribe to see EXIF info for this image (if available)


If you look at the telephone wires at the top of the photo you can see that they are not straight, but are bowed up. Here is the same image processed from the raw through PhotoLab:

8mm-Setting2.jpg
Subscribe to see EXIF info for this image (if available)


Of course it is much wider, since it is the full 8mm image, but when processed through PhotoLab the telephone wires are much straighter. they are a bit bowed, but in the down direction, which is what you would expect.

The take-away seems to be that the auto de-fishing done by the E-M1.2 is good, but it is not as good as that you would get from PhotoLab. I do not mention Lightroom because I was never able to get it to give me the full image after correction; it kept insisting on lopping off the sides to retain the original image size ratio.

Having said all of this I am still very fond of the lens. It looks good, feels great, the focus ring is as smooth as silk and the auto de-fishing gives me a chance to see the image in the EVF close to how it would look after processing. The auto de-fishing is still very good, and in many of my test shots I could not see any difference between the in-camera and the out-of-camera adjustments, but if you shoot raw as I do, then you would probably want to end up using the raw files anyway.

I still intend to do a bit more testing as I want to determine what FOV of the 3 settings give you. What I have been able to determine is that the narrow setting gives you an image that is just slightly narrower than 9mm, I think about 9.5mm, and I believe that the medium setting gives something like 7.5mm, but I need to do some additional testing and get out my trig tables to get more exact figures.

Of course, if anyone already knows, they could give me the figures and save me some work ...
 

ArizonaMike

Mu-43 Top Veteran
Joined
Jan 19, 2018
Messages
817
Location
Arizona, USA
> Could you show us that last image using the settings 1, 2 and 3, and also not defished?

Yes.

Narrow:

FE-Marrow.jpg
Subscribe to see EXIF info for this image (if available)


Medium:

FE-Medium.jpg
Subscribe to see EXIF info for this image (if available)


Wide:

FE-Wide.jpg
Subscribe to see EXIF info for this image (if available)


Not Defished:

OLY00001-8-NotDefished.jpg
Subscribe to see EXIF info for this image (if available)


A couple of comments.

1) All of the OOC auto de-fished jpgs are cropped, and even the wide jpg loses quite a bit of information. I suspect that Olympus just cropped the wide where it assumed the side distortion from the de-fishing became too obvious.

2) All 3 jpgs are the normal pixel size, 5184 x 3888, so the operation is not done in the camera just based on cropping or, if it is, the camera interrelates and supplies the missing pixels. Cropping each jpg image 1:1 yields different images.

3) The distortion correction done by the camera is really pretty good, but not as good as that done by LR or PhotoLab. I took three images this morning of a screen on the side of a house. Compare the OOC jpgs with the PhotoLab image, judging by the de-fishing only.

Narrow:

Screen-Narrow.jpg
Subscribe to see EXIF info for this image (if available)


You can see the barrel distortion easily enough.

Medium:

Screen-Medium.jpg
Subscribe to see EXIF info for this image (if available)


Wide:

Screen-Wide.jpg
Subscribe to see EXIF info for this image (if available)


It seems to me that the medium setting is the most accurately de-fished, but compare those with the output from PhotoLab on the raw image:

Screen-Raw.jpg
Subscribe to see EXIF info for this image (if available)


When I bought the lens I made the basic assumption that Olympus knew best how to de-fish their own images, so I thought that the distortion correction would be better than from any 3rd party software. I have changed that opinion, based on some basic testing.

4) The firmware settings provide the user with a pseudo zoom functionality for this lens. That is, you can select any of the three settings are get different zoom values. Of course it is all software, but in looking at the different images I did not see any of the artifacts I normally see when I resize an image, so it seems almost as good as a normal zoom lens. If I keep this lens I may well sell my 9-18 as unnecessary, given this "zoom" functionality. If this lens gives me an equivalent of, say 10mm, at its greatest "zoom", then my 12-40 can take over from there and I do not need to worry about missing the focal lengths provided by the 9-18. Otherwise I may get the Laowa.

> This may answer your question about why Lightroom is cropping your output

I spent some time this morning playing with LR to try to stop it from just throwing the side information out the window and I did find a way to do that. However it is not in the crop section, but in the Transform section. I had to adjust the scale so that the entire image was visible, then use the crop tool to crop it the way I wanted. It works, but it seems like a lot of work when they could just have provided a checkbox somewhere on the UI that would have told it to not keep the original image ratio.
 

Growltiger

Mu-43 Hall of Famer
Joined
Mar 26, 2014
Messages
2,199
Location
UK
Thank you for all that work. It shows exactly what I wanted to find out.

The results are disappointing and I am amazed that the camera is not doing a proper defishing. It is doing a sort of 80% defish followed by some cropping.

I would not be happy with this lens as a substitute for a normal rectilinear ultrawide. Every photo with a straight line in it would have to be defished properly in post processing.
 

ArizonaMike

Mu-43 Top Veteran
Joined
Jan 19, 2018
Messages
817
Location
Arizona, USA
Thank you for all that work. It shows exactly what I wanted to find out.

The results are disappointing and I am amazed that the camera is not doing a proper defishing. It is doing a sort of 80% defish followed by some cropping.

I would not be happy with this lens as a substitute for a normal rectilinear ultrawide. Every photo with a straight line in it would have to be defished properly in post processing.
I am happy that I could be of some help.

Personally I have mixed emotions about these results.

On the one hand, Olympus provides a moderate amount of distortion correction, and that shows up in the EVF which allows me to have enough information to properly frame the photo. It de-fishes well enough to preserve horizontal to vertical proportions and, unlike my Samyang, PhotoLab has correction information for it. It does not have that for the Samyang and I have to pretend that it is an Olympus 8mm. And it provides 3 image settings which allows me to use it as a variable focal length fisheye lens, and since I shoot raw anyway I would probably end up doing the processing on the fishy raw images. And, of course, it is an f/1.8 lens.

On the other hand the distortion correction it does should be better. I can see no reason that Olympus itself can not do as good a job correcting this image as PhotoLab or Lightroom. It is expensive and Olympus does not seem to provide any way to de-fish the raw image. I do not mean in the camera, but in the Olympus Viewer. And it seems to me that Olympus ought to provide some functionality to allow the user to apply the same distortion correction to the raw that is applied to the jpg. Not in the camera, but on the computer. There should be some Olympus tool that says "correct the distortion of the raw based on the correction in the jpg", or something like that.

It is a nice lens. It feels good on the camera, the balance is good, the finish is excellent and I do like the auto focus. It would be even nicer if there was a function button on the lens.

I have not decided if I will return this lens or not. I do a lot of landscape shooting since we travel to the National Parks all the time. This lens should make that easier, but I am not sure it actually does, and I am considering if the Laowa 7.5mm would not be a better (and less expensive) option.
 

Growltiger

Mu-43 Hall of Famer
Joined
Mar 26, 2014
Messages
2,199
Location
UK
I will be sticking to my current choice of 9-18 when travelling and Oly 7-14 when not. And the 12-40 is often perfectly adequate.
 

JanW

Mu-43 Veteran
Joined
Dec 11, 2014
Messages
248
Location
Nieuwerkerk a/d IJssel, Netherlands
Real Name
Jan
What about corner sharpness?
Although the Photolab perspective correction is better there is hardly any sharpness left in the corners.
Mayby that is why Olympus are a bit conservative on their correction and made a compromise between correction and sharpness.
How does the corner sharpness compare to the 9-18?

Since you already have the Samyang for fish-eye perspective I wonder if you would be better off with the Laowa 7.5 for rectilinear UWA work?
 

ArizonaMike

Mu-43 Top Veteran
Joined
Jan 19, 2018
Messages
817
Location
Arizona, USA
What about corner sharpness?
Although the Photolab perspective correction is better there is hardly any sharpness left in the corners.
Mayby that is why Olympus are a bit conservative on their correction and made a compromise between correction and sharpness.
How does the corner sharpness compare to the 9-18?
Interesting question. I had not thought to check that; I was so surprised to see the barrel distortion that I thought about little else.

I will check.

Since you already have the Samyang for fish-eye perspective I wonder if you would be better off with the Laowa 7.5 for rectilinear UWA work?
I bought the fisheye to use it for rectilinear panoramas and I had no desire to make fisheye images to keep as fisheye images. If I decide that the 8mm is the wrong lens for me due to the issues with de-fishing I will return it, sell the Samyang and buy the Laowa. I want a pano lens, fisheye or not.

I do not have the sense of art that makes people take interesting fisheye images. I was an engineer before I retired and I just don't have that creative sense.
 
Joined
Feb 15, 2014
Messages
12,349
Location
Southwestern Utah, USA, Earth
Interesting question. I had not thought to check that; I was so surprised to see the barrel distortion that I thought about little else.

I will check.


I bought the fisheye to use it for rectilinear panoramas and I had no desire to make fisheye images to keep as fisheye images. If I decide that the 8mm is the wrong lens for me due to the issues with de-fishing I will return it, sell the Samyang and buy the Laowa. I want a pano lens, fisheye or not.

I do not have the sense of art that makes people take interesting fisheye images. I was an engineer before I retired and I just don't have that creative sense.
I think you are selling the engineer's short. This is coming from an engineer's son, and a pretend engineer while I am at work.

I am following this thread with much interest. I also just received the Oly 8 mm PRO. I have had the Rokinon 7.5 for 4 years now, and I use it very often. I personally like the fisheye image, especially with lots of distortion (dial to 11). I have also been de-fishing images from the Oly and I see that there is potential.
 

ArizonaMike

Mu-43 Top Veteran
Joined
Jan 19, 2018
Messages
817
Location
Arizona, USA
I think you are selling the engineer's short. This is coming from an engineer's son, and a pretend engineer while I am at work.
I apologize if I offended you, but I was referring to myself when I mentioned a lack of artistic creativity, not engineers in general.

I am following this thread with much interest. I also just received the Oly 8 mm PRO. I have had the Rokinon 7.5 for 4 years now, and I use it very often. I personally like the fisheye image, especially with lots of distortion (dial to 11). I have also been de-fishing images from the Oly and I see that there is potential.
This was always a means to avoid stitching panos together for me. I am often in the US National Parks and was looking for a way to create very wide images to try to capture the landscape. Earlier this month I was at Arches National Park and Natural Bridges National Monument, both in Utah, and in a couple of weeks I expect to be at Bryce Canyon National Park, Zion National Park and, perhaps, Kodachrome Basin State Park, also in Utah, and all of those places just cry for UWA images. The 8mm purchase was for that sort of thing.
 

ArizonaMike

Mu-43 Top Veteran
Joined
Jan 19, 2018
Messages
817
Location
Arizona, USA
What about corner sharpness?
Although the Photolab perspective correction is better there is hardly any sharpness left in the corners.
Mayby that is why Olympus are a bit conservative on their correction and made a compromise between correction and sharpness.
How does the corner sharpness compare to the 9-18?

Since you already have the Samyang for fish-eye perspective I wonder if you would be better off with the Laowa 7.5 for rectilinear UWA work?
I went back to look at the corners of the images, but am not sure which posted images you are referring to. I checked the screen images, but they seem to be OK at the corners. I checked the rest of the images but did not notice any real issue with the corners other than where the image has been stretched for the distortion correction.

Perhaps you could specify exactly what you meant? Which 8mm images? The narrow? The wide? And which corners? Remember, the closer you get to the edges of the images the more distortion there will be due to the projection changes.
 

JanW

Mu-43 Veteran
Joined
Dec 11, 2014
Messages
248
Location
Nieuwerkerk a/d IJssel, Netherlands
Real Name
Jan
Perhaps you could specify exactly what you meant? Which 8mm images?

I mean the last image of your original post.
The gravel in the bottom corners is completely smeared out while te original fish eye image (in your second post) is very sharp.
The 1MKII jpegs are also sharper in the corners than the Photolab image.

And I completely understand your remark about engineer vs creative sense. I'm technician and my 13 year old daughter often tells me to be more creative :dash2:
That's why I look at the corners of your images I suppose....

Jan
 

ArizonaMike

Mu-43 Top Veteran
Joined
Jan 19, 2018
Messages
817
Location
Arizona, USA
I mean the last image of your original post.
The gravel in the bottom corners is completely smeared out while te original fish eye image (in your second post) is very sharp.
The 1MKII jpegs are also sharper in the corners than the Photolab image.
That is an interesting observation.

I do not know much about the internal workings of the algorithms for de-fishing an image, but I assume each one works slightly differently. Dxo hides what it does, but if you use Lightroom you can go to the Transform panel and change the scale to see what is actually being done. If you do you get this:

LR-DIstortionCorrection.jpg
Subscribe to see EXIF info for this image (if available)


On the other hand if you use PTLens and do the same thing you get this:

PTLens.jpg
Subscribe to see EXIF info for this image (if available)


And it is easy to see that the two are different by looking at the bottom right at the shadow of my taking the photo. They are both distorted, but in different ways.

I suppose your observation means that Olympus pays more attention to the distortion at the edges of the crops they make while PhotoLab and Lightroom, which are virtually identical, pay less attention there, but perhaps more attention elsewhere. I never thought about that, so I am grateful that you were so observant. I think I need to do more testing. It may well be that I would be better off accepting the Olympus corrections than those from PL or LR, although Olympus only corrects the jpgs, not the raws.

As I mentioned in another post, if I keep the 8mm I expect to use it for landscape photos, not architectural photos, so the slight remaining distortion after the in-camera corrections is really not important to me.

Thanks for mentioning this.
 

ArizonaMike

Mu-43 Top Veteran
Joined
Jan 19, 2018
Messages
817
Location
Arizona, USA
Mayby that is why Olympus are a bit conservative on their correction and made a compromise between correction and sharpness.
Your post made me really curious about the differences between the de-fishing algorithms used by different software, so I decided to take a look at PhotoLab vs the Olympus Viewer. I thought that perhaps Olympus used a similar de-fishing algorithm in their Viewer software to the one used in camera to produce the raws. The result was very interesting.

Here are both images:

PhotoLive vs Olympus Viewer.jpg
Subscribe to see EXIF info for this image (if available)


If you look on the right you will see 3 circles and an arrow, and they show how much difference there was in how Olympus de-fished the image compared to PhotoLab. As you observed earlier, they appear to do far less distortion at the edges than PhotoLab does, and that probably accounts for the sharper edges.

To be clear: in all of the differences you see in the marked sections, the Olympus Viewer image is the "inside" the PhotoLab one. The center of the image matches between the two, but the edges definitely do not.
 

JanW

Mu-43 Veteran
Joined
Dec 11, 2014
Messages
248
Location
Nieuwerkerk a/d IJssel, Netherlands
Real Name
Jan
Thank you for sharing all of this with us.
Very nice to see the differences between the algorithms!
I don't have the funds for an em1mkII or an Oly 8mm but these observations are probably also valid for the samyang 7.5 mm. With that lens on my em10 there is of course no in-camera (or OV3) defishing so it will depend on the right defishing software.

At the moment I have no UWA so maybe the 9-18 will be the best bet for a start in wide angle landscapes.

Jan
 

Latest threads

Links on this page may be to our affiliates. Sales through affiliate links may benefit this site.
Mu-43 is a fan site and not associated with Olympus, Panasonic, or other manufacturers mentioned on this site.
Forum post reactions by Twemoji: https://github.com/twitter/twemoji
Forum GIFs powered by GIPHY: https://giphy.com/
Copyright © Amin Forums, LLC
Top Bottom