3.0 What is a Panosphere
My entire thesis was built upon the concept of a panosphere, a term coined by Industrial Light & Magic (ILM), and my desire to replicate what was accomplished in the first Avengers movie in 2012. It's hard to believe that movie came out just before my freshman year of college! I was captivated by the fact that the entire city was recreated digitally. A small crew was sent to photograph large sections of the city using a technique that captures extensive panoramic photos forming a sphere, known as a panosphere.
To learn more about their process watch this feature
I aimed to achieve what they did for NYC, but on a smaller scale - the RIT campus.
Spoiler alert: Due to time constraints and the vast scope of my project, this goal was never fully realized.
Background
Industrial Light & Magic (ILM) created several thousand panospheres, or panoramic photo spheres, that they stitched together to recreate New York City for the movie Avengers (2012). For the movie, they built a basic low-poly 3D city and projected the panospheres onto their 3D geometry. Flying a helicopter over the streets of New York City is not permitted, so they had to improvise and develop a creative way of flying through the city. This is a common challenge for many films, where something cannot be done practically, so the visual effects team is brought in to solve the problem (ILMVisualFX, 2013).
Several panospheres were created for my thesis, and then compared with the photos taken of the chrome ball.
A panosphere is essentially a panoramic photo stitched together from a single point in every direction. In virtual reality, the user would be at the center of the photo or where the camera was located on set. This concept was still relatively new at the time, as VR has gained more traction in recent years.
The panospheres are stitched together using the software PTGui. This requires multiple photos to be taken in such a way that each photo overlaps enough information with the others; identifiable features, such as trees and buildings, help match the photos together. Creating a panosphere is a lengthy process, especially with the tools I used on set. While there are ways to automate this process, I had to do it the old-fashioned way as a budget-conscious college student.
These panospheres took anywhere from 3-5 hours to complete, which may seem like a lot, but I was also shooting these to be HDRI. The time consumed in this process could be significantly reduced with the right equipment.
Similar to the Chrome Ball, the images here can be used for lighting information. The way they are treated in post is a different process and they can be used for more than just a reflection map. Learn more about that in the Chrome Ball VS Panosphere post.
The Setup
At the time, I was more focused on the actual process rather than documenting it. This photo shows my tape markers, which were used to help align the entrance pupil.
The camera needs to be placed carefully on the tripod so that the entrance pupil of the lens revolves around the same point. The entrance pupil will be slightly different for each lens used, but typically there is a red dot on the lens signifying where the entrance pupil is located. This is where the z-axis is located, and the camera needs to spin around this point.
Some tripods might not allow the camera to be placed where it needs to be. This was the case with the tripod used in this project. Ideally, the tripod should enable the camera to slide back and forth enough for the entrance pupil to align with the center of rotation on the tripod head. However, the tripod I used did not allow for this, so some improvisation was necessary.
After identifying the entrance pupil, I needed to take some fishing wire with a weight tied to the bottom and attach it to the tripod in a way that aligned it with the entrance pupil.* As I moved the camera around, I could ensure that the weight lined up with the center marked on the floor.
The Process
Photos should be taken at intervals of approximately 5 to 15 degrees around the z-axis in a circle.
Next, repeat the process with the camera tilted 5 to 10 degrees upward.
Continue until the camera is fully pointing upward.
Afterward, repeat the same process going below the initial axis (or the starting axis).
The following post provides an example of photos from a successful stitch.
Post production
If the photos are not taken properly, PTGui may struggle to stitch them together. In PTGui, you can edit control points to fix some of the stitching errors, but this can be time-consuming depending on the number of photos imported for the stitch. The software I used was the trial version, which included a watermark on every export. I planned to purchase the software if this workflow proved viable.
Since I incorporated HDRI into these panospheres, I repeated the same process for several stops above and below normal exposure. This provided high dynamic range information about the scene. Once all the photos were taken, they could be merged similarly to the chrome ball through Photoshop's automation feature.
Afterward, each photo is ready to be imported into PTGui. Once PTGui completes the processing, you can adjust the photo. This is where the stitch needs to be fine-tuned to eliminate any visible artifacts. The length of this process depends on the number of photos imported, but I found it to take a long time.
The first panosphere I stitched together had issues due to an insufficient overlap between the photos. The software allows users to make modifications and changes to which photos are stitched. The first panosphere required significant modification and didn't turn out as expected.
Panosphere stitched together in Photoshop
While trying to upload this photo, I realized how large the file was – [3575x1117] pixels at 300 ppi, over 100 inches wide. I modified it to fit this website, but as an example, I've cropped into this large photo document to show its true size and detail.
Conclusion
One of the challenges of this project, especially with the goal of capturing panospheres all over campus, was finding enough time with the right weather while balancing other classes and projects. It's worth noting that it took ILM an entire summer to capture New York City through panospheres with their small crew.
Due to the extensive amount of time required for this process, I had to stop pursuing it further and focus on other aspects of the project. However, I did manage to capture several successful panospheres and was able to compare them with the photos of the chrome ball I had taken. I believe I spent an appropriate amount of time on this section, going through the process, stitching the photos, and having something to compare with the chrome ball.
I wish the time involved in this process wasn't so extensive, but I think the results were quite impressive! I discuss this further in the post comparing Chrome Ball VS Panosphere.
Notes:
The fishing wire cannot be seen because these are the photos I used for the actual stitching process; so after lining things up, I held the fishing wire and weight out of frame.
