Picture Code

[MikeFromMesa]

The type and number of images are exactly the same as for the B5/ASP procedure. If you've calibrated lenses in ASP then the procedure is the same, but the tool is different.
F_P

No, I have not calibrated lenses in ASP.

Did you read that link? So for a 18-55 I took shots at 18, 21 24, 28, 35, 45, 55. In ASP you have to use some external s/w to calculate the (a,b,c) parameters for each focal length and insert the results in the lens DB (a text file). ASP interpolates for intermediate focal lengths. PN is similar. You use the line straightening tool on each of these images and then "train" and PN will also interpolate for intermediate focal lengths. That's my understanding and experience. What's the problem? The only hard part is to select the right kind of wall / building to use as your target. Do not use brick walls.
F_P

Yes, I did read the link but, as I understand what you are saying, it is not necessary to do any s/w calculations since all I need to do is use the straightening tool and then "train" PN at the various focal lengths. That certainly seems a lot easier to me.

I re-read the link with what you have said in mind and it now seems like a very easy process. Find straight lines to photograph, start taking images from about 3 meters or so away, take then every 10% (for the edges) and 15% (for the rest) of the various focal lengths, straighten the images and then "train" them. For a 17-35mm lens that would seem to be 17, 19, 21, 24, 27, 30, 32 and 35 and that is probably more than I need to do. I assume the more straight lines in the image the better and that I should be straightening all of them.

It is too dark to do a good job now but I will try this tomorrow. I hope that this process helps take care of the distortion in my lens when processed by PhotoNinja and I wonder if it will have any effect on the vignetting I see with my UWA when it is wide open. I assume the distortion correction will push some of the image out of the frame and that may have an effect. Also, when I sent one of these images to PictureCode to ask them why the vignetting adjustments in PN do not correct the vignetting in the image, one of the developers responded that perhaps PN, since it shows more of the image than specified by Canon, is displaying too much at the edges. It will be interesting to see if the lens distortion adjustments take care of the vignetting.

Thanks for the response.

[ferdinand-paris]

The hard part is finding the straight lines to photograph. It's not as easy as you think. Took me a while. I use a wall on the side of a building with large black tiles. You need at least one line that goes most of the width of the image, and you want it about 2/3 from the centre of the image to the top. You have to position it there while holding the camera horizontal in all dimensions. Ideally (if you're a perfectionist like me) you get a second line about 1/3 of the way up, plus another line 2/3 of the way from the centre of the image to the left (or the right), which goes most of height of the image. There's nothing magical about the top vs the bottom, but I strongly suspect that you'll find it easier to get lines at the top of the image.

I am sure that someone will chime in and say that only the first line is necessary, and that's probably true in many cases, but some forms of distortion can be complex. Did I mention that I tend to be a perfectionist?

F_P

[MikeFromMesa]

The hard part is finding the straight lines to photograph. It's not as easy as you think. Took me a while. I use a wall on the side of a building with large black tiles. You need at least one line that goes most of the width of the image, and you want it about 2/3 from the centre of the image to the top. You have to position it there while holding the camera horizontal in all dimensions. Ideally (if you're a perfectionist like me) you get a second line about 1/3 of the way up, plus another line 2/3 of the way from the centre of the image to the left (or the right), which goes most of height of the image. There's nothing magical about the top vs the bottom, but I strongly suspect that you'll find it easier to get lines at the top of the image.

I am sure that someone will chime in and say that only the first line is necessary, and that's probably true in many cases, but some forms of distortion can be complex. Did I mention that I tend to be a perfectionist?

F_P

I am confused (not such an unusual feeling for me).

You say to use the side of a building and then say that the line should be about 2/3 from the center to the top of the image. The side of a building would be a vertical line, not a horizontal line, and so I would think it would be about 2/3 from the center to the left or right side of the image, not the top or bottom. What am I missing here?

My plan was to take some square pillars holding up a gazebo roof near here and make some shots of them with one pillar on the left and the other on the right of the photo. But those would then extend vertically, not horizontally and you are talking about horizontal lines. So, as I said, I am somewhat confused ...

[ferdinand-paris]

A picture is worth 1,000 words, or in the case of ambiguous words like mine perhaps 10,000. I will try to post an illustration. There used to be something online, but I can't find it now.

In the meantime, I will try to clarify this statement "You need at least one line that goes most of the width of the image, and you want it about 2/3 from the centre of the image to the top". You are shooting in landscape and you need a horizontal line that goes from the left of the image to the right. Going all the way to both edges isn't necessary, but it should go most of the way. The position of this horizontal line should be towards the top of the image. The comment about 2/3 was a measure of the height of the horizontal line. These lines need to stay away from the image centre, where there is little or no distortion. So you want it above the image centre, and closer to the top of the image than the centre, like twice as close.

That may or may not be clearer. Sorry for any confusion.

F_P

p.s. Here is an example: http://epaperpress.com/ptlens/calTargets.html
This has the horizontal line closer to the top that I use, but I'd also use a slightly lower horizontal line, plus one of the verticals towards the left or right.

[MikeFromMesa]

A picture is worth 1,000 words, or in the case of ambiguous words like mine perhaps 10,000. I will try to post an illustration. There used to be something online, but I can't find it now.

In the meantime, I will try to clarify this statement "You need at least one line that goes most of the width of the image, and you want it about 2/3 from the centre of the image to the top". You are shooting in landscape and you need a horizontal line that goes from the left of the image to the right. Going all the way to both edges isn't necessary, but it should go most of the way. The position of this horizontal line should be towards the top of the image. The comment about 2/3 was a measure of the height of the horizontal line. These lines need to stay away from the image centre, where there is little or no distortion. So you want it above the image centre, and closer to the top of the image than the centre, like twice as close.

That may or may not be clearer. Sorry for any confusion.

F_P

p.s. Here is an example: http://epaperpress.com/ptlens/calTargets.html
This has the horizontal line closer to the top that I use, but I'd also use a slightly lower horizontal line, plus one of the verticals towards the left or right.

OK. So if I understand correctly I should be able to turn the camera 90 degrees (to portrait instead of landscape mode) and shoot one of the vertical pillars, making sure that I keep the camera level at 90 degrees and that the pillar edge is about 2/3 of the way "up" the image from the center. I can then turn the image back to landscape when I put it on my computer and "viola", I should have the proper image? Or perhaps the camera will understand that the image has been rotated and the whole exercise will be a waste of time.

Yesterday I took an image of a set of vertical pillars with one on the left of the image and one on the right. I then straightened both lines as well as the lines at the top of the structure and the resulting image, before cropping, looked as though someone had pushed in both the left and right sides substantially. I tried to represent the shape in this post but when I "previewed" it the processing removed all of the spaces and it looked nothing like the original shape I drew so I hope my explanation is clear.

I will see what this new image does, but what surprised me was that when I "trained" the software it took the image as representing both the focal length and focusing distance. I understand why, but that means that I will need a lot of images to cover the focusing distances (although I assume the closer distances are more important). In any case it will be entertaining.

Another question is whether any adjustment setting will work for both sides of the image. That is, if I use straight lines toward the top and left, will the software understand that the adjustment needs to be symetric and also adjust the bottom and right? And, if not, how do I combine two image settings to take care of that. My (limited) experience with PhotoNinja training is that a new image for the same aperture and focusing distance will replace an old one so I cannot just use two images. I have been trying to find something where there was a line on the left and on the right and at the top and finding something like that has not been easy.

[ormdig]

I have found the new version of Photo Ninja giving unwanted color cast to it's automatic highlight recovery when activating a new image.

Have you wound down the "Strength" slider of "Color Recovery" in the "Color Correction" tab? That should fix it.

I was getting false colour highlights sometimes (along with tons of detail mind, unlike ASP) so I've adjusted the default slider setting to 70, and I've barely had an instance of it since doing so: and the rare time it's happened since, taking the slider down a bit further fixes it.

Thank you. I set the default at "medium" and several similar images activated without the false color. Problem solved.

[ferdinand-paris]

OK. So if I understand correctly I should be able to turn the camera 90 degrees (to portrait instead of landscape mode) and shoot one of the vertical pillars, making sure that I keep the camera level at 90 degrees and that the pillar edge is about 2/3 of the way "up" the image from the center. I can then turn the image back to landscape when I put it on my computer and "viola", I should have the proper image? Or perhaps the camera will understand that the image has been rotated and the whole exercise will be a waste of time.

That's not what I was trying to say. All the lines need to be in the one image. Why don't you just try with one line as per that PTLens page and see how you go? My idea of three lines (in the one image), is a bit of overkill. I will try to find time to post an illustrative image.

I will see what this new image does, but what surprised me was that when I "trained" the software it took the image as representing both the focal length and focusing distance. I understand why, but that means that I will need a lot of images to cover the focusing distances (although I assume the closer distances are more important). In any case it will be entertaining.

I noticed that too, but I think you only need to do that if you are an ultra-perfectionist. I haven't bothered so far.

Another question is whether any adjustment setting will work for both sides of the image. That is, if I use straight lines toward the top and left, will the software understand that the adjustment needs to be symetric and also adjust the bottom and right? And, if not, how do I combine two image settings to take care of that. My (limited) experience with PhotoNinja training is that a new image for the same aperture and focusing distance will replace an old one so I cannot just use two images. I have been trying to find something where there was a line on the left and on the right and at the top and finding something like that has not been easy.

My understanding is that these approaches to correction assume that distortion is symmetric. As noted above, all the lines for one focal length need to be in the one image. You are right that finding the right target is hard. If you just use the one horizontal line approach it's easier. Start with that.

F_P

[MikeFromMesa]

That's not what I was trying to say. All the lines need to be in the one image. Why don't you just try with one line as per that PTLens page and see how you go? My idea of three lines (in the one image), is a bit of overkill. I will try to find time to post an illustrative image.
F_P

As it turns out I found what looks like the perfect place to train the software. I had never really thought about it before but my house has an open roofed area in the backyard for sitting and staying out of the sun. There are vertical walls going straight up to the roof and wooden roof supports across the top. This forms an almost perfect upside down U (with a stright top) and I took a series of images at different distances yesterday with both my Canon 24-105 and my Tamron 17-35 making sure I kept the walls and roof close to the edges of the image.

When I began the walls and supporting roof beam looked like graceful curves and I used them to "train" PhotoNinja. By the time I got to the last images there was nothing left to do. The software had already corrected the curved walls and roof so that they were as straight as the tool lines. It really seems to work like a charm. I will now need to go and find some lines that are, perhaps, more geometrically perfect to test what I have done but it is clear to me from just looking at the images that this simple exercise has helped enormously.

I am very grateful for your help. This fixes one of the few issues I had with PhotoNinja (no way to automatically correct lens distortion when my wide lenses were fully open). Now if you only had a suggestion for how to do batch processing ...

Thank you again.

[herman]

Now if you only had a suggestion for how to do batch processing ...

Sorry for breaking in in your chat.
Batch processing is easy: just have some patience.
They are working on it

[herman]

Some time ago I learned from the Photo Ninja tutorials that Photo Ninja has an extremely smart way to interface with LightRoom.
When you call PN as external editor from within LR an intermediate file <imagename>-edit.tif is created by LR.
When PN finds that file in it's inbox it starts searching for the corresponding original raw file and opens it.
When you are done converting the raw you overwrite the intermediate -edit.tif file with the PN rendering and continue your work in LR with a file which has been converted by an external raw converter.

I thought it might be a good idea to have a similar workflow in ASP, so I asked the PN developers to add the ASP naming convention of intermediate tif files (which is _edit.tif) to the PN parsing mechanism.
The kind people at PictureCode implemented this in version 1.0.4, which is currently available as pre-release.
I tested the mechanism tonight, it works like a charm on my system.

This may be helpful to ASP users who want to integrate PN into their workflow without sacrificing tools available in ASP.

[ferdinand-paris]

Batch processing is easy: just have some patience.
They are working on it

Not in 1.0.4. Do you have any idea how much patience we are likely to need?

F_P

[herman]

Batch processing is easy: just have some patience.
They are working on it

Not in 1.0.4. Do you have any idea how much patience we are likely to need?

F_P

No, I don't have an ETA. But then I did not ask it .....

[afx]

Not in 1.0.4. Do you have any idea how much patience we are likely to need?

You can ask them to put you on a mailing list for that announcement.

cheers
afx