Panorama over-processing

[ms.fuentecilla]

After getting the panorama processing near perfect in the last iteration of the official programme the new beta seems to have reverted to over processing resulting in bleaching out of detail that is subsequently difficult or impossible to adjust satisfactorily. It would be a pity were this to be carried through into the next main iteration of AP. I tested this by setting up the same panorama sequence in both. The difference is significant.

[Rick G]

16 minutes ago, ms.fuentecilla said:

After getting the panorama processing near perfect in the last iteration of the official programme the new beta seems to have reverted to over processing resulting in bleaching out of detail that is subsequently difficult or impossible to adjust satisfactorily. It would be a pity were this to be carried through into the next main iteration of AP. I tested this by setting up the same panorama sequence in both. The difference is significant.

Could you make the photos available for us to test? I will also make a test with my own photos to see if it is the beta or if something else might be going on

[ms.fuentecilla]

Thank you for your interest. It will be difficult to make the clips available. The compilation uses I think 6 or 7 frames at 12Mp and my internet speed in Spain is desperately slow. They are jpegs. They were made on automatic exposure in Dutch winter sunlight on a Pentax WG3, which causes some variations from frame to frame. Those get ironed out impressively in general. The point really is that AP main made an excellent compiled image and the beta adjusted the light areas too far, so I wasn't sure whether there had been some amendments in the intervening pipeline. Since making the test I have made other panoramas with the beta from the same archive with no problems. The best I can do at present until again in Holland is to present screenshots of the results, if that helps.

I would like to make it clear that I find this module excellent. The only warfare I have with AP is selection refinement!

[harrym]

That's interesting as I spend a fair amount of time processing panos!

Are you saying that if I process the same files in the current release version and the current beta version I'll see a difference?

Regards

[ms.fuentecilla]

Making more trials in the beta with the same frames I find inconsistencies. Some results are acceptable and similar to those from the release version. The panorama concerned is of part of a historic landscape garden, a small lake surrounded by trees in winter, one half in shade the other in sun. On some results the trees in sun bleach out; on others they don't. It is a difficult subject perhaps with a quite high contrast range, but fairly typical of landscape panoramas. If this involves only making different versions then perhaps it is no great problem, not a major issue. AP's superb panorama module facilitates quick processing. Interesting though. And an incentive to use my cameras with RAW output with flexible post processing perhaps. In part the jpeg compression might be responsible, though the original sunlit frame is well balanced.

I'm not saying that you will get a difference HarryM, only that I have but it might be interesting if you tried?

Much to my surprise I have managed to upload the frames involved. They were hand held using my PentaxWG3

[Sean P]

Hi ms.fuentecilla,

I've tried your three images attached and both 1.6.2.93 and 1.6.3.98 are giving me the same results - with the exception of the transparent background (due to another issue).

 

See below:

https://www.dropbox.com/s/m89x56ft3sw2zpe/Export_1.6.2.95.png?dl=0

https://www.dropbox.com/s/rabw6mor5kruddp/Export_1.6.3.98.png?dl=0

[ms.fuentecilla]

A few problems with dropbox but I did manage a download to examine. 1.6.3.98 is similar to the better results I got from retries. My original concern was that alterations might have been made to algorithms but it seems not and this might be an inconsistency in my system. I repeat that I find this module of AP superb (which implies no criticism of its other qualities). I struggled for years trying to get similar results from Photoshop! Thank you Sean for your interest. I think this query can be closed, don't you?

[ms.fuentecilla]

I revisited this panorama with the new release version of AP and the result is as near perfect in the processing as could possibly be expected. There is no bleaching, excellent balance.

[ms.fuentecilla]

This is fine with horizontal panoramas but with vertical panoramas severe bleaching of skies happens when they are processed the normal way up. When the procedure inverts the image it is much more successful. This suggests preconceptions about processing skies perhaps? Also the green patches are very apparent most in the over exposed areas.

[ms.fuentecilla]

Problems remain with the panorama module. It is unable to handle high contrast RAW images and the output RBA has insufficient flexibility to control the resulting image resulting in bleached out highlights. Overcoming this entails exporting adjusted  component images to TIFF files before processing as a panorama. This is a tedious workflow. It does seem to me that either the internal adjustments of the process need further consideration or the final output needs to be an HDR of some sort rather than something the equivalent of JPEG in which most of the detail in highlights is submerged if a professional status is intended.