My Work - The Process
On the 21st of November I was booked for a corporate photography job back near when I originate in Cumbria. This was just a day job and I had scheduled the whole weekend back at home. I rang my grandma after returning from the commission and arranged to shoot the very next morning.
The shooting process was simple and fairly efficient. I ran my grandma through what I was planning on doing and got her to briefly talk about the memories she could think of around the room. For the narration, I didn’t want it to feel forced or scripted so every now and then I had to stop her from continuing the story and tell her to remember it for when I was recording. In hindsight I could have simply set the audio recorder off during these moments to allow these more pure tellings of the stories but I was conscious about having my voice and others in the recording. Although, this could have possibly been for the better, creating a natural dialogue between us both about the memories we both share, but that’s not what I envisioned.
Once I’d got a general feel for the objects being talked about and the key areas of the room, I spent a little while organising the room to hold these objects in a fairly evenly distributed fashion. In doing this, I was channeling the original 360 room tour uses for this kind of photography. It presented the room not as an immediately inhabited state, but more of a piece of art in itself. Similar to Tracy Emin’s bed project, the arrangement of objects, furniture and the general mise en scene plays a huge character in itself. The oddities of seeing how my grandma lines up the remote controls for the TV and her excessive use of picture frames and blankets is something that (although not directly related to my grandmas memory) could bring up ideas of peoples own experiences, relatives and habits.
I set up my tripod in the middle of the room and ensured I used a spirit level to make sure the head was level, I then did the same with my cameras in built spirit level - working using a gyroscope within the body - and span the head a whole 360 to ensure It stayed this was throughout the panorama. The kit I used was My Nikon D7000 with a Tamaron 10-24 at it’s widest setting. I used an ISO of 400, an aperture of f6.3 and a shutter speed of 1/30th of a second. This exposed the majority of the room (especially areas near the conservatory and windows) a little on the over exposed side. This was the look I was going for though, retrenching the work of Will Pearson. I shot in RAW and knew I could balance this exposure out better in post by retaining the shadows and bringing down the highlight areas as opposed to exposing for the highlights and having to try and drag detail back in the shadows. This often doesn't work and ends up with washed out shadows and an un natural and often unpleasing contrast. From this I set about shooting my panorama in normal fashion. Shooting wide open with at 10mm meant that I managed to get a lot of the room into each frame, I matched up easily rememberable points in each frame with the last (edge of chair on far right - next frame - far left) I ensured I gave plenty of leeway for this for when blending the images in post. This would hopefully mean that I had maximum coverage per photograph and minimise the amount of work in post. I rotated 6 times to create the complete panorama. I then panned the camera up and repeated the process to get the upper third of the room and once more for the lower. I also shot a single frame looking straight up and straight (or as close as possible) down. Due to the wide angle lens however, I would still manage to get the full coverage. To get this properly, I’d have to invest in a specialist panoramic head for the tripod that allows the camera to be raised above the plate situated on the tripod legs with an extra vertical plate to attach the camera. This allows the lens to point straight up/ down and often has degrees for precise rotation throughout. I’m hoping that the way I am shooting this time would be sufficient but depending on the outcome and my desire to continue the experimentation, I may wish to try one of these out.
After shooting the full panorama, I brought my grandma back into the room and set up two audio recording devices. As a ‘scratch track’ (so to speak) I used my DSLR and simply filmed my grandma traditionally. I also set up a shotgun microphone plugged into my iphone to record more detailed audio. Once I had both of these recording I sat down and asked my grandma to recite the information that she had talked to be about before hand. I often had to prompt her to remember certain things she had discussed but tried to remain silent otherwise and quickly cut myself off when she started talking. Although the microphone I used is directional. There would still be bounced audio and obvious discourse between us two so I wanted to minimise that as much as possible for more seamless editing. I repeated this exact process for both of the rooms I shot.
Post processing:
This has been the most difficult (as is often the case) of the whole process. I started off fairly simply by dragging all my images into Adobe Lightroom. From here I started with a fairly neutral image (one containing the most contrasting artefacts such as windows and shadows) I then edited this until the images histogram was a little more spread out and the image took on this HDR esc look so familiar with panoramic work. I then highlighted the rest of the images in the series and used Lightrooms handy ‘sync’ function. This syncs the develop settings with all of the selected images and because I had shot them all at the same settings they would all turn out with a consistent look throughout. It wouldn’t rid the image of its shadows or highlights but flatten them out and retain the detail in these areas. I then exported these images as high quality JPEGs with little sharpening into a separate folder.
I have used the photoshop action merge to panorama in the past to create panoramas and so I didn’t need much introduction to the process. However, upon using the script, it seamed to get a little confused with this level of files and detail. I tried selecting different settings such as the output from automatic to dedicated spherical and even perspective but all seamed to get confused and ended up stitching the photographs incorrectly. The only way I could get the script to work was to only add in one row of the images. I inserted the straight on images from the series and ran the script again and It worked fairly well with little artefacts or corrections. I was uplifted by this but this only left me with a very tight panorama and would not work for a full 360 VR experience.
I thought that if I could keep doing this as separate photographs (straight on - upwards tilt - down) and then stitch these separate panoramas manually then I’d be onto something (obviously not as efficient as I would have hoped) but this was not the case. Once I deviated from the standard view and introduced frames with less points of interest for the script to stitch with, it fell apart and couldn’t create a cohesive image. It often ended up in a circular form or taking on an S curve shape. It became evident that this program was not going to work for the intensive workload that this project required.
I took to the internet to look for alternatives on image stitching software/ plugins and settled on a program I had tried in 6th form - Autopanno Gigga available through the French company, Kolor. This software is regarded as the industry standard for panoramic image making. It’s been used on countless groundbreaking projects including the worlds largest panoramic image - In2white. http://www.in2white.com/ This image is comprised of over 70,000 images shot over 35 hours using a specially designed motorised head. As a finalised product, the image “weighs in at a staggering 365 gigapixels.” and needed specialist software to host the image online. This is a fully interactive image with the ability to zoom in and see individual climbers and the details of the mountain range almost miles away.
If that alone isn’t testament to the capabilities of the software, then I don’t know what is. I decided to give it a go and upon downloading the software, I was a little intimidated. This is obviously a specialist program that needs time and dedication to perfect. The interface Is very technical, looking like the interface for a CAD 3D program as opposed to image processing software. I started reading a few posts on forums online on how to best use the software but this only confused me more. Being dyslexic, I often find it hard to digest information in this way and see the way this translates to my own application. Instead, I watched a few tutorial videos mainly on YouTube. This ability to see cause and effect of the processes and the situation of different features of the program made a whole lot more sense and made my grasping of the program much quicker. I of course referred back to theses often more in depth articles and forum threads on the software but in connection with this more visual tutorials. This video was probably the most helpful in my understanding - https://www.youtube.com/watch?v=1ntRoV97k9o - very condensed and straight to the point.
Firstly, I loaded the exposure adjusted exports from Lightroom into the program. I then select all of these images and runt hem through a 3rd party application through the Autopanno Gigga program to remove geometric artefacts and adjust the from radius on the images to help the program in finding the stitching points. These are then exported into the same folder as the Lightroom exports for future use and are place into a separate project file within Autopanno. I then adjusted the panorama settings to output as a spherical 360 image and to adjust the amount of control points available for the smoothest output. I also optimised the render settings for best quality and compatibility by rendering as a TIFF format with no compression. I then ran the program’s script. This was an automated process and i selected the files to be automatically rendered and saved during this to save time and my need to be sat at the computer so I could continue with other elements of the project. This took roughly an hour and a half to complete but it was worth it. The finished product was a lot smoother and exact than the outcome of the photoshop script. I was very pleased with the results and although it still took a couple of attempts to get a result as seamless as I’d have wished, it was still a lot less work intensive than the Photoshop rout. I could have used further processing tools within the program to adjust each image layer (similar to working on a photoshop document) to perfect these outcomes but that required hours and hours of work that I simply did not have. Finding time to look at these tools outside of the course is something i’m very much excited about and maybe would wield a better result in the future.
As you can see from this finalised output from Autopanno Gigga, the result is not without it’s faults. Some of the lines along the walls have been skewed and can not be matched perfectly. This may be due to the fact I used a wide angle lens and the geometric distortion caused by this. This is part of the learning curve however and is something that my tutor discussed in our group tutorials this week. Taking more photographs with smaller increments might have also alleviated this somewhat but also by using a longer focal length such as a 35mm or 50mm which has less distortion attached to it. Of course this would have slowed down the workflow by introducing maybe double or triple the amount of photographs and the time taken to physically take these but could be worth it in the end.
This all being said, I have a panorama that contains 360 degrees of a room and matches up perfectly. These small artefacts can be minimised in photoshop and are a small price to pay for (what I consider) an impressive image already.
I then took this uncompressed TIFF (1.11gb’s in size) back into photoshop and adjusted these issues to the best of my abilities. I also masked over the tripod legs seen in the shots lower half. This created my final image which I then slightly adjusted the saturation and levels of till I had an image with a little more punch. It took till seeing this finalised image to feel the need to introduce this slight contrast and saturation because viewing all 30 odd images as one is a different experience than viewing them individually. This is what’s exciting about panoramas and VR photography, the sense of scale and grander alters how we read the image even down to a tonal sense. Once I was happy with the image I exported it as another uncompressed TIFF.
I tested the images effectiveness using an app I downloaded for my phone called Looksee. This is a VR app for displaying photographs and allows for control over the image in 3D using the phones gyroscope and the users movement. I made sure that all areas of the photograph matched up and there was no blank or stretched areas. This included the poles of the image.
From here I imported the image into adobe premier where I’d set up a project best for internet usage. due to the fact I wasn’t using moving image in my video, the fps didn’t matter too much. I made sure to set the frame size at 1920x960 for the highest quality output that fits the panoramic shape. I had the audio bit rate set to 44000 for sharp and uncompressed audio giving a good tonal range. As I was rendering out the other panoramas I set about editing the audio files to cut out dead space and unnecessary talking etc. I then exported this as one MP3 track for each video and dropped it into the timeline along with the corresponding panorama. Then it was just a matter of making the time the panorama was on screen match the time of the audio. I then exported the video using the h.264 codec under a preset for YouTube 1080 playback. This added black bars to the top and bottom of the video but this won’t matter once the video is uploaded to youtube and set to the 3D viewer inside the headset.
In order for YouTube to understand and project the image as a 3D viewable video, I needed to alter the files metadata. To do this I used a very simple program called Spatial Media Metadata Injector. All that’s required is to upload the video into the program and select 2 out of the 3 options “My video is spherical (360)” and “My video is stereoscopic 3D (top/bottom layout)”. The third option is for if the video contains 3D stereoscopic sound which did not apply to my project at present. Once this has been exported from the program it’s set to be viewed in most VR players. Youtube being my chosen one. To do this, is as simple as uploading any regular video to the site and YouTube will automatically display the video to best present the 3D aspects. For desktop, this means allowing the user to drag the photo in all axis using the curser or using the up/down/left/right controls on the top right of the video. For mobile and VR headsets, the video is controlled using the mobiles in built gyroscope and the users movement.