I shot first few HDR’s with the Sigma 10-20mm superwide lens, but it was not wide enough to shoot panoramas efficiently. So then I bought the Samyang 8mm f3.5 manual-focused fisheye lens with horizontal field of view of 180 degrees. This made the task alot easier, though, the Samyang was pretty cheap lens, so from something it is always off. The Sigma had smaller aperture by 0.5 stops, but where Samyang takes advantage in aperture-size, it takes back in optical quality. Samyang has considerably more chromatic aberration. As like with Samyang, the Sigma also draws better when shooting few stops below maximum aperture. Usually the developers of the lenses, especially cheaper ones, trade-off the image quality for maximum aperture, so if the scene allows it, it is better to stop down the lens for f5.6-8 for example.
For HDR-panorama compilation, I use software called PTGUI, which is really easy to use, and yields excellent results. Though there was few difficult steps in using it. I wondered that why my images kept getting such high values (when stitching the images together, PTGUI informs the final stitching result as a average pixel distance from one image to another. Higher values mean more parallax and more visual errors in the image). Then I found out from the PTGUI FAQ-section, that the Sigma is a superwide lens but aspherical in perspective, and for some reason PTGUI reads the lens EXIF-data wrong and thus gives wrong results and unsuccessful stitching. So when using Sigma 10-20mm superwide lens, check off the tickbox which reads the EXIF-data, and switch the lens type to “Rectilinear” then PTGUI understands it correctly and gives appropriate results.
And as for Samyang, it does not feed lens information into the camera body, so the settings have to be manually adjusted for PTGUI. Choose in the lens settings “Full Frame Fisheye”. And from the crop factor of 1,6 (for Canon DSLR’s) and insert the focal length as 9mm, although it is really 8mm. Then PTGUI will give correct results.
As for preprocessing the panorama, PTGUI can detect bracketed exposures and fuse them into an HDR-image while stitching, but I have found out that the workflow is much easier to manage, if the individual shots are compiled into HDR-images for example in Photoshop, and then the HDR’s stitched together in PTGUI. I usually compile the HDR-files in Photoshop, and then bring them into PTGUI for initial stitching. If the stitching is not successful, it usually is that the HDR-files come a bit too dark from the Photoshop and so PTGUI can’t find appropriate stitching points. Usually then I go back to Photoshop, open the HDR-files and pump-up the exposure with the value of two for each image. Do not worry the apparent overexposure, Photoshop is only showing a part of the dynamic range, and for example going to the Image -> Adjustmens -> HDR Toning and choosing from the dropdown menu “Highlight Compression”, you can see that there is plenty of dynamic range to retain all the color-intensities.