Announcement

Collapse
No announcement yet.

The Garden at the V and A

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    The Garden at the V and A

    I know, a bit soon after the last one (although the one before that was over a year ago), but I'd really like to know if people think this is okay? I messed about with it quite a bit, then undid a lot of it to get here... brain's a bit fried now (still it took my mind off how much the roofing guy is charging me):

    #2
    Re: The Garden at the V and A

    You never fail to amaze me with some stunning panoramas.
    Canon 1DX, 50D, EF500 F4.0 L, EF100-400 f/4.5-5.6L I , EF100-400 f/4.5-5.6L II, EF70-200 f/2.8L II, EF180 f3.5L Macro, EF 24-105 f/4L, EF17-40 f/4L, EF2.0X III, EF1.4X III, 430EX II, MR-14EX...

    Comment


      #3
      Re: The Garden at the V and A

      That is just stunning work, how many shots make this up?

      Regards Paul

      Comment


        #4
        Re: The Garden at the V and A

        Another great pano and with so many moving people I can't begin to imagine how much editing was needed

        Cheers,
        Joihn

        Comment


          #5
          Re: The Garden at the V and A

          still it took my mind off how much the roofing guy is charging me
          London huh, That'll be all those chimney sweeps and nannies with brollies dancing on the roof that's causing the damage.
          Canon 1DX, 50D, EF500 F4.0 L, EF100-400 f/4.5-5.6L I , EF100-400 f/4.5-5.6L II, EF70-200 f/2.8L II, EF180 f3.5L Macro, EF 24-105 f/4L, EF17-40 f/4L, EF2.0X III, EF1.4X III, 430EX II, MR-14EX...

          Comment


            #6
            Re: The Garden at the V and A

            Originally posted by pelliott1954 View Post
            That is just stunning work, how many shots make this up?

            Regards Paul
            Five, or strictly four and a bit as I only used a teeny bit of the "Down" one. I shot the 8-15 at 12mm and portrait mode in NSEWD (not strictly North, the trick is to fit stuff you'd rather not have to stitch into one shot and go from there).

            I generally find if you shoot at 8mm (with a circular Fisheye) and FF you get double coverage everywhere (with NSEW shots, opposing pairs aimed either up a bit or down a little), so stitching and choosing which people (or general moving objects) you want in a shot is more flexible and the final Pano is around 90% of the pixel count of your sensor (so a 20MP sensor gives about an 18MP spherical pano - it's so low as the circle only uses a small part of the 3:2 sensor and you throw half the data away due to the double coverage).

            The reason for shooting 12mm and Portrait orientation is it's as far as you can go to shoot NSEW and have it work. Now you use almost all the sensor area with a modest overlap and so the pano is around 270% the pixel count, hence this pano shot with a 50MP camera is actually a bit over 134MP (15,836 x 7918, also an 877MB TIF file, which makes the c. 10MB limit for Viewat somewhat exciting).

            You can zoom out further and take more photos, but stitching gets harder and it's a case of how much the extra detail adds as now I'm using a 5Dsr you get very good detail in the 12mm panos (and IMHO the 8mm ones, the Chapter House pano allows a good zoom on the stained glass). The main issue is you have to shoot at multiple vertical angles with greater than 12mm focal lengths, which adds a whole pile of fun to the stitching and the shot count goes up rapidly.

            One part I like is choosing what goes into the pano, partly through choosing when and where to shoot and partly by masking.. This one has an optional toddler playing in the water over the far side of the pond, but I decided to leave him out (by choosing data from the overlapping image instead) as I had tried quite hard to shoot round the youngsters playing at the pool.

            Other different panos:

            Mild HDR: http://www.viewat.org/?i=en&id_aut=7...0&pag=1&sec=pn
            Each shot bracketed +/- 2 stops.

            One of my pushing my limits ones, 32 images using a 14mm f2.8 shot freehand (which was a bit silly I guess), until the Millennium Bridge one it was easily the most time I took stitching (IIRC about 225MP):


            Only 26 images with the 14mm, plus using a bit of string to help with the alignment:


            Examples of how you can actually contribute to the image artistically, rather than being a DIY Google Street Maps:


            (The boats + train + people one was quite challenging but I think came out okay.)

            Note none of my 44 panos on Viewat used a tripod or a pano head... which allows shots in places that would be very hard otherwise and helps with the stuff at your feet due to the lack of a tripod (provided you remember to move your coffee cup, my first try at the British Museum one had a Pret cup on the floor, actually my first try at the V and A had the weight on the end of the cotton I was using for positioning, as I forgot to move it for the last tilted down shot).
            Last edited by DrJon; 12-04-2017, 21:31.

            Comment


              #7
              Re: The Garden at the V and A

              In case it's of interest here's the two "tilted down a little" images from my first try, where I forgot to move the weight out of the shot for the last one (SooC JPEGs)...
              VandA-try1.jpg
              (It's a little dark as I have to avoid highlight clipping, also the Sun was behind a cloud - I was a bit torn between the "Sun behind a cloud" approach where everything was a more even brightness and "Sun not behind a cloud" where the good stuff would look better. I went with the latter. I shot four sets of images, two in Sun and two in Cloud.)
              Last edited by DrJon; 12-04-2017, 15:40.

              Comment


                #8
                Re: The Garden at the V and A

                Post-processing Technique for the V&A image:
                (Update - "Simple version" hints added for beginners.)

                (1) Make a Lightroom profile using a Color Checker passport shot at the scene.
                Simple version - ignore this.

                (2) Process Raw files, with the custom profile, in Lightroom 6.9 (or 6.10?) to get as much shadow/highlight detail as possible and taking care to set the purple/green fringing controls carefully. Output 16-bit TIFs using exactly the same settings for all five images.
                Simple version - use good quality JPEGs, maybe SooC.

                (3) As the South shot was quite dark and has some dark areas then do noise reduction in Neat Image and save as a 16-bit TIFs. (This is quite unusual, Lightroom NR is usually enough - I actually realised later that some of what I thought was noise was actually very low brightness detail seen through the door.) Use the exact same noise profile and Neat Image settings for all images
                Simple version - ignore this. Actually I've done noise reduction only a couple of times in total.

                (4) Load into PTGui Pro, set cropping circle. Set a vertical line on each of the NSEW images. Set a horizontal line on the N image around the horizon level. Set control points between the D image and each of NSEW images. Mask out the weight and any bits of me in the images. Mask out any people I don't want or look less good than the alternative people (in the over-lapping areas).
                Simple version - just do one vertical line (within one image) and mask out stuff to suit.

                (5) Align images. Optimise. Optionally optimise the exposure.

                (6) Adjust masking. Add a few more control points to help the Down image (which was having difficulties). Optionally pick the pano centre in the viewer.
                Simple version - no down image so no issue.

                (7) Make pano as a 16-bit TIF. View in a Pano viewer.
                Simple version - or a low compression JPEG. I use the free FSPViewer64 to look at images.

                (8) Adjust masking and add control points based on how it appears, rinse and repeat from (5) a few times.

                (9) When it's close enough load into Photoshop. Fix any stitching errors you've seen (by the time this step comes along they shouldn't be that hard to do). View in a Pano viewer. See more stitching errors and fix them too. (Sometimes come back to this point after you've posted the Pano online when you see yet another error.)
                Simple version - ignore this.

                (10) When the Pano looks good sort out the look. I actually have six adjustment layers in Photoshop.
                Simple version - ignore this.

                (11) Save final pano as 16-bit TIF.
                Simple version - ignore this.

                (12) Save preview image for ViewAt using a custom crop setting.
                Simple version - note from here on are ViewAt specific.

                (13) (Ab)use the Photoshop save-for-web control to get a JPEG that is as close as possible to the ViewAt 10,485,760 byte file size limit.

                (14) Use PTGui Pro to make a .MOV file for devices (e.g. iPhone, iPad) that like this format for Panorama viewing.

                (15) Use PTGui Pro to make six faces for a Panorama (as ViewAt does tend to choke on at-limit JPEGs, so having a six-face pano is usually a good option). Zip up the six JPEGs. See how close to 10,485,760 bytes you are then adjust JPEG ratio, save again, ZIP again, repeat until get the best possible result.

                (16) Write up the blurb for Viewat.

                (17) Post Pano on ViewAt.

                Then hopefully don't spot any big errors, ever. (Note I keep a record of the compression ratios for the 10MB files and the MOV, actually in the file-names, so they are easier to duplicate later.)

                VandA_files.jpg

                Simple version - shooting tips.

                My advise for people starting out is to consider a circular fisheye lens, i.e. either 8 or 7.5mm on a FF camera (adjust per crop factor for others). It's much easier and so more likely to give good results.

                Equipment: Camera with circular fisheye lens, a length of cotton and a metal bolt, plus a ColourChecker Passport (err, or just choose Sunny as the White balance, or 5600K).

                I tied a piece of cotton around the front end of the lens (as the nodal point is somewhere around there, although strictly it's the "no-parallax point" you want) and tied a weight to the other end of the cotton (a metal bolt - although after banging it into the lens I wrapped it in soft paper and put it in a plastic bag and changed to a rubber band to connect to the lens). I picked my location and used the cotton+bolt to keep the same position and height to take 4 photographs at 90 degree intervals. I tilted the camera up 10 degrees for the 1st and 3rd shots and down 10 degrees for the other two. (Okay, more "a bit" than exactly any number of degrees.) I used manual white balance (set from a grey card, actually a ColorChecker passport in my case - these days I shoot the colour patches too) plus used manual exposure at f8 (I just metered around me in A mode and picked a good compromise). I tried to set the focus correctly to have 1m to infinity in focus (which is just past the 1m mark on my 8-15mm lens) and left that in manual focus too.

                I use PtGui Pro (commercial software), with the free SmartBlend plugin, for stitching. The free Hugin software is probably worth a look. IMHO the non-Pro version of PTGui isn't, as missing the masking option is, I think, a show-stopper. I chose SmartBlend as the blender and Lanczos16 as the interpolator. I ran all the tabs in Advanced mode. I also ticked all the viewpoint boxes in the optimiser (well, except the top one, which doesn't apply).

                Tips:
                * Try to get anything you don't want to stitch all in one shot, e.g. on the Chapter House all four stained glass windows are from one shot (well, a set of three bracketed shots).
                * Shoot a "Down" shot as well (just freehand) as it can help with later problems. (I'd say don't use it initially, as it needs a lot more work to make it fit in.)
                * If the ceiling is very complicated (e.g. my British Museum Atrium pano) shoot up as well, aiming to get it all in one shot.
                * Pull the weight away just before shooting the last down shot, so you have the ground without it.
                * Try to avoid having lots of moving people around where you will want to put a join.
                * If people are moving in one particular direction rotate against them, so you won't have the same person appear twice.
                * If in bright Sun move around between shots (you, not the camera, that stays still) so you can mask out your shadow and still have data from another shot to cover it. Look for shadows to hide in.
                * Look for a mark on the ground so you know where the weight is for when you move it accidentally, as you will, often.
                Last edited by DrJon; 15-04-2017, 14:29. Reason: Added "Simple version" stuff.

                Comment


                  #9
                  Re: The Garden at the V and A

                  Thanks John This detailed and extensive explanation of the process is a perfect example of why we should always revisit posts we've already viewed and commented on. Had I not I'd have missed this - now all I have to do is read and understand it

                  Cheers,
                  John

                  Comment


                    #10
                    Re: The Garden at the V and A

                    Oh and in answering the "how many images" question I really should have included:

                    Note the joins shown are the nominal joins, SmartBlend will do quite a bit of messing about to get a nice blended join (other blending options are available, SmartBlend is particularly slow so most people probably use the built-in Blender, but I find SmartBlend works well for my style of hand-held shooting).
                    Attached Files
                    Last edited by DrJon; 14-03-2018, 21:30.

                    Comment

                    Working...
                    X