I have a set of raw (.RAF) images which I wish to process in Hugin. I forgot to shoot with jpeg as an option, so I have to export these raw files, after editing them in dt , to approximate the look of a Fujifilm jpeg. I have a concern (based on earlier editing) that I will end up with inconsistent luminance and chrominance across the set of images, even when I use the copy/paste of the history stack from 1 image which I choose to specify as the ‘master’ image. Furthermore, I don’t think that the spot color mapping feature of the color calibration module is going to give me consistent color: the manual says this feature is for ‘consistently lit’ images, which mine are not: these images were taken across a 180 degree (at least) arc across the view, traversing from the sun being behind the camera to the camera looking directly at the sun.
What process should I use to get consistent jpegs that do justice to the Fuji camera, to give a good result from Hugin?
What Fuji camera?
Do you have access to Fuji X Raw Studio?
Are you talking about a very large number of images, or would it be feasible to re-process the RAFs into JPGs inside the camera?
Camera: X-T30
Fuji X Raw Studio: I guess I can install it on my windows PC
18 images total - I was not aware that I could re-process within the camera - is this what x Raw Studio is for? I’ll have to read up about it; it would be a super easy approach. Thanks for the suggestion.
Sadly, no. the ref. to a page number requiring 3 digits (thankfully only decimal) gives a clue as to why. I had tried, but my forehead could take no further punishment - let alone the damage to the veneer on my desk top. I did try to read a proper printed bookl: I purchased a much touted (by Fujifilm UK) book on the X-series. It was so woefully written - aside from the numerous technical errors and the utterly confusing style - that I gave up with it, turning to YouTube instead.
And that’s a place where people who should never have been given command of a video camera, a microphone, a monotone style of speech, a seriously restricted vocabulary, a tenuous grasp of grammar and diction, and a deep commitment to the belief that arm-waving, finger-pointing and grotesque leering constitute effective communication, tend to congregate. It is a vivid demonstration of why we need the Darwin Award.
But, thanks to your reference to Fawlty Towers (q.v.), I have how read the manual as well as installed and used X Raw Studio. That all worked wonderfully well. My work with Hugin less so - far. far less so. Perhaps I should read the manual …
For consistent processing, make sure the same amount of exposure correction is applied to all images. Do not try to equalise them in darktable: Hugin will read the EXIF exposure info, and perform the equalisation.
Use an exposure value that avoids clipping; apply that to all the images, even if taken with different exposure settings; that way, relative exposure relationships will be maintained.
White balance
Ensure all your images use the same white balance. Depending on your chromatic adaptation settings, do one of the following:
legacy (white balance) only: apply the same white balance multipliers to all images
modern (white balance = camera reference + color balance rgb): make sure the source image has the illuminant set to custom; apply the same illuminant (settings of the color balance rgb module) to all the images.
Tone mapping: don’t
Do not apply any curves (such as filmic, sigmoid or base curve): export linear data.
Export your images, process them in Hugin; take the image from Hugin, adjust exposure as if it were a raw file, apply any curves, contrast enhancements and so on.
At least that’s how I work. I’ve found that if I have a large number of overlapping images, the fewer I actually use in Hugin, the lower the chance of ghosting. Therefore, in Hugin I disable all images that do not cause gaps to be shown. Also, I have never had luck with XDR output from Hugin.
That video is very interesting - and directly relevant. I’m surprised I missed it. I see it dates from December 2021. In it Aurelien states that it is experimental, available only to those who compile their own version of dt - which is, beyond question, not me. However, my version of dt (4.2) appears to have these very same functions in the exposure and color calibration modules. These functions appear to directly address my concern about inconsistent color mapping, but it is not clear to me how ‘measured’ values can be applied as ‘corrective’ values to a selection of images, other than through selective copy and paste of the history stack - which I fear will introduce un-intended changes. I’ll have to give a try.
I think this is why he made them persistent. You will see that resetting the module doesn’t clear them so you need to manually do it… if you paste append I think it should be fine… but I think because they are persistent you can just use the picker in each image and have it apply that setting…
I don’t recall the step by step in the video but i think he demonstrates how to apply the process… good luck…
Ah! The significance of that was lost on me, while being made an obvious feature of the function. I should have paid more attention, as this use of persistence within the spot color mapping is (I think) unique within dt. Thanks for reminding me.
If you don’t have any success with dt and hugin, you can try the free Image Composite Editor from Microsoft. It is not open source and I’m not shilling for a commercial company, but credit where it’s due, it is free to use and does a very good job. You can input RAW files and it will tone map them and blend them automatically. You can then export as a TIFF and finish your editing in darktable.
For less work on the exported TIFF, I have had success using the spot exposure and spot colour mapping tools in dt to prepare JPEGs first, and then use ICE to blend them together.
Its caught a lot of people that forget to set it back to 50 0 0 in CC module… Then everytime they use the picker for a spot WB they get a weird cast…. Bit of a gottcha if you forget…
I’ve never had an issue with this - make sure you use selective copy, and tick all the boxes. I also find it’s a good idea to have lens correction enabled for pano images, to remove vignetting and distortion.
Thanks for this tip. I quite agree with your assessment, which is worth more than 1/-; it’s at least 1/9d, if not half a crown, in old, real, ‘rattles in your pocket’, money. It’s a pity that MS have made it so difficult to find ICE; the only (minimally reliable) source I could find was (for the benefit of subsequent readers) via dpreview.com/forums/threads/4551999 - at the bottom of that thread, where a dropbox URL is given.
I was greatly surprised to find that ICE - dating from 2008 I guess - was able to handle my RAF images which, for my camera, post-date ICE by about 10 years. How does it do this, aside from being able to read the embedded jpeg with the RAF, I suppose?
I know I’m strictly ‘trailing-edge’, but before starting this ‘walk-about’ I had no idea that this panorama stitching technology is at least 2, if not more, decades old. Why no recent, FOSS, activity on this function? Have we already delivered the ultimate solution? If so, the ‘usability’ aspect seem to be well disguised in some of the offerings …
Indeed you have and I thnk you for taking the time to do so. I haven’t yet had the time to absorb all your advice - but I will do so shortly. In the meantime, the big problem I have is in understanding what a ‘stack’ is in Hugin and how to get Hugin the stitch my images…
On my last photographic ‘event’ I took two sweeps, of 9 images each, across the view, one above the other in order to not lose parts of the roof-tops or the ground in the resultant cropped panorama. I even used a 14 mm (21mm in 35 mm equivalent, I think) lens for this, with a tripod mounted camera. I don’t understand how to tell Hugin how to position these images relative to each other. I thought that by manually setting the control points I would then be telling Hugin explicitly how the images related to each other. That quickly became all too big a task. In any event, the resultant panorama appeared as if all the images had been laid on top of each other, in what I would describe as a single ‘stack’.
At this time I have decided that I must break a long -standing tradition of falling asleep while reading the manual. This is a challenge: the manual is written from the viewpoint of describing/specifying individual function points (which is truly soporific), not from the task perspective. The YouTube videos are generally too superficial, aside from being tremendously irritating to watch. But I will persevere…
Do you have enough overlap between the images? Hugin will try to look for similar pieces in the photos (anchor points). If you do not get enough, or they are of bad quality, stitching won’t work.
Here’s Aurélien Pierre’s tutorial. I actually apply the white balance, instead of doing it after stitching (that somehow never worked for me).
