Wednesday 28 February 2018

How I Created a 16-Gigapixel Photo of Quito, Ecuador

A few years ago, I flew out to Ecuador to create a high-resolution image of the capital city of Quito. The final image turned out to be 16 gigapixels in size and at a printed size of over 25 meters (~82 feet), it allows people see jaw-dropping detail even when viewed from a few inches away.

I’ve always thought that gigapixel technology was amazing since I first saw it around 8 or 9 years ago. It combines everything that I like about photography: the adventure of trying to capture a complex image in challenging conditions as well as using high tech equipment, powerful computers, and advanced image processing software to create the final image.

I’ve been doing this for a while now, so I thought that I would share some of my experiences with you all so that you can make your own incredible gigapixel image as well.

The Gist

The picture was made with the 50-megapixel Canon 5DSR and a 100-400mm lens. It consists of 912 photos with each one having a .RAW file size of over 60MB. To create the image a robotic camera mount was used to capture over 900 images with a Canon 5DSR and 400mm lens. Digital stitching software was then used to combine them into a uniform high-resolution picture.

With a resolution of 300,000×55,313 pixels, the picture is the highest resolution photo of Quito ever taken. This allows you instantly view and explore high-resolution images that are over several gigabytes in size.

Site Selection

The first step in taking the photo is site selection. I went around Quito and viewed several different sites. Some the sites I felt were too low to the ground and didn’t give the wide enough panorama that I was looking for. Other sites were difficult to access or were high up but still not able to give the wide panoramic view that I was looking for.

I finally settled on taking the image from near the top of the Pichincha Volcano. Pichincha is classified as a stratovolcano and its peak is over 15,000ft high. I was to access the spot via a cable car and it gave a huge panoramic vista of the entire city as well as all the volcanoes that surround Quito.

The only drawback that I saw to the site is that I felt that it was a little too far away from the city and I didn’t think that people would be able to see any detail in the city when they zoomed in. To fix this situation I decided to choose a site a bit further down from the visitor center. That meant that we would have to carry all there equipment there (which isn’t easy at high altitudes) but I felt that it would give the best combination of a great panoramic view and be close enough to the city for detail to be captured.

The Setup

The site was surrounded by very tall grass as well as a little bit of a hill that could block the complete view so I decided to set up three levels of scaffolding and shoot from the top of that. There wasn’t any power at the site since it was on the side of a volcano so we had to bring a small generator with us.

I ran extension cords from the generator up to the top of the scaffolding where it powered the panorama head, as well as my computer. I didn’t plug in the camera in because I would be able to easily change the batteries if they ran out.

Atmospheric Conditions

Anything that affects the light rays on their path to the camera’s sensor will affect the ultimate sharpness of the image. Something that is rarely mentioned is the effects of the atmosphere on high-resolution photos. Two factors are used to define atmospheric conditions: seeing and visibility.

Seeing is the term astronomers use to describe the sky’s atmospheric conditions. The atmosphere is in continual motion due to changing temperatures, air currents, weather fronts and dust particles. These factors are what cause the star images to twinkle. If the stars are twinkling considerably we have “poor” seeing conditions and when the star images are steady we have “good” seeing conditions.

Have you ever seen a quarter lying on the bottom of a swimming pool? The movement of the water makes it look like the quarter is moving around and maybe a little bit blurry. Just as the movement of water moves an image, atmospheric currents can blur a terrestrial image. These effects can be seen in terrestrial photography as the mirage effect, which is caused by heat currents and also as a wavy image due to windy conditions. It’s interesting to note that seeing can be categorized according to the Antoniadi scale.

The scale is a five-point system, with 1 being the best seeing conditions and 5 being the worst. The actual definitions are as follows:

  1. Perfect seeing, without a quiver.
  2. Slight quivering of the image with moments of calm lasting several seconds.
  3. Moderate seeing with larger air tremors that blur the image.
  4. Poor seeing, constant troublesome undulations of the image.
  5. Very bad seeing, hardly stable enough to allow a rough sketch to be made.

(Note that the scale is usually indicated by use of a Roman numeral or an ordinary number.)

Visibility: The second factor that goes into atmospheric conditions is visibility, also called visible range is a measure of the distance at which an object or light can be clearly discerned. Mist, fog, haze, smoke, dust and even volcanic ash can all effect visibility.

The clear high altitude air of Quito made for some amazing visibility the day of the shoot. The only things that affected it that day were a few small grass fires in the city. The Cotopaxi volcano was also giving off smoke and ash but it didn’t seem to be a problem since it was blowing away from the city. There also weren’t any clouds in the sky which made it so that the exposure wouldn’t be affected by any clouds blocking out the sun.

Equipment

Camera: I decided to use a 50 megapixel Canon 5DS R. The 5DS R is an amazing camera that is designed without a low-pass filter which enables it to get amazing pixel-level detail and image sharpness.

Lens: A Canon 100-400mm f/5.6 II lens was used to capture the image. Several factors went into the decision to use this lens such as size, wight and focal length. The 100-400mm was small and light and would allow the robotic pano head to function with no problems. It also has a good focal length of 400mm with would allow for some nice detail to be captured.

I thought about using a 400mm DO and 400mm f/2.8 but each had its own drawbacks. The 400mm DO didn’t have a zoom and I wanted to be able to change the focal length for different types of captures if I had any problems and the 400mm f/2.8 was too big and heavy to be used properly in the pano head. I have a Canon 800mm f/5.6 which I would have loved to have used but it was also too heavy to be used with the robotic pano head (humble brag).

Another interesting factor that went into my decision to use the 100-500mm f/5.6 is that the diameter of the front lens element was small enough so that atmospheric distortion wouldn’t be too much of a problem. I have spent a lot of time experimenting with astrophotography and the larger the front lens element is the more atmospheric distortion or “mirage effect” will be picked up resulting in a blurring of the photo.

Pano Head: I used a GigaPan Epic Pro for the image capture. The GigaPan is an amazing piece of equipment which automates the image capture process. The GigaPan equipment is based on the same technology employed by the Mars Rovers, Spirit, and Opportunity, and is actually a spin-off of a research collaboration between a team of researchers at NASA and Carnegie Mellon University.

To use a GigaPan you first need to set it up for the focal length of the lens that you are using. You then tell it where the upper-left-hand corner of the image is located and where the bottom-right-hand corner of the image is. It then divides the image into a series of frames and automatically begins scanning across the scene triggering the camera at regular intervals until the scene is completely captured.

There are several other brands of panorama heads out there including Nodal Ninja and Clauss-Rodeon but I like the Gigapan the best since it is automated, simple and reliable. The GigaPan is also able to be connected to an external power source so the battery won’t run out during large image capture sequences.

Computer: I didn’t think that the memory card would be large enough for all the images to be stored on it especially since I was going to be making multiple attempts at capturing the image. I decided to shoot with the camera tethered to a MacBook Pro via Canons EOS Utility. This software not only allowed me to write the images directly to my hard drive, it also allowed me to zoom into the image in live view to get critical focus. Just in case something went wrong I simultaneously wrote the images to an external hard drive as a backup.

Camera Settings

Aperture: I set the aperture to f/8. This was done for a couple of reasons. The first was to increase the resolution of the image. Although the Canon 100-400mm f/5.6 II is a very sharp lens shooting wide open, stopping down the lens a little bit increases its sharpness. Stopping down the lens also reduces vignetting, which is a darkening of the edges and corners of the image.

Although the vignetting is minimal on the lens, I have found out that even the slightest amount of vignetting on the frame will result in dark vertical bands being shown in the final stitched image.I didn’t want to stop down the aperture too much because I was worried about diffraction reducing the resolution of the image.

Focal Length – I shot at 400mm so I could capture as much detail in the city. I could have used a 2x teleconverter but there was so much wind at the site that I was afraid that the camera would move around too much and the image would come out blurry.

ISO: I shot at an ISO of 640 due to all the wind at the site. I knew that using a high ISO would increase my shutter speed and reduce the chance of vibrations from the wind blurring the photo.

Shutter speed: All of these factors combined gave me a final shutter speed of 1/2700.

RAW: I shot in .RAW (actually .CR2) to get the maximum resolution in the photos.

Live View: I used the cameras live view function via Canons EOS Utility to raise and lock the mirror during the capture sequence. This reduced the chance of mirror slap vibrating the camera.

Gigapan Settings

The Gigapan has a lot of different settings for the capture sequence of the images. One can shoot in columns from left to right or in rows from the top down and left to right or any combination thereof. I choose to shoot the image going across in rows from top down going from left to right. Even though the image capture sequence would only take an hour or so I have found that shooting in this sequence makes for a more natural looking image in case of any change in lighting conditions. I also included a 1-second pause between the GigaPan head moving and the trigger of the camera to reduce any shake that may have been present from the pano head moving.

Image Capture

I had to go at it a few times but the final image was taken with 960 photos with each one having a .RAW file size of over 60MB.

Image Processing

Two Image Sets: Each day of the shoot presented itself with different problems. One day the city was clear but the horizon and volcanoes were obscured with clouds. On another day the horizon was totally clear. I decided to create two different image sets and combine them together to make the final image. One large image set was used for the clear sky and volcanoes another image set was used for the city.

Pre-Processing: For the horizon and volcanos I selected an image that I felt represented an average exposure of the sky into photoshop and corrected it to remove any vignetting.

For the image set of the city found an exposure of the city and color corrected and sharpened it to the way I wanted it before bringing the images into the stitching software. I recorded the image adjustments that I made and made a photoshop droplet with them. I then dragged and dropped all the files onto the droplet and let it run, automatically correcting each image of the photo sequences. It took a long time but it worked.

Autopano Giga: After the images were captured I put all of them into Autopano Giga. Autopano is a program that uses something called a scale-invariant feature transform (SIFT) algorithm to detect and describe local features in images. These features are then matched with features in other frames and the images are combined or stitched together. The software is pretty straightforward but I did a few things to make the final image.

Anti-ghosting: Autopano has something called an “anti-ghosting” which designed to avoid blending pixels that don’t match. This is useful for removing half cars or half people that could show up in the image due to the movement of objects between frames.
Exposure blending – Just in case of any vignetting or differences in the lighting I used the exposure blend function in the software to even out the exposures and make a nice blend.

.PSB: .PSB stands for Photoshop Big. The format is almost identical to Photoshop’s more common PSD format except that PSB supports significantly larger files, both in image dimension and overall size.

More specifically, PSB files can be used with images that have a height and width of up to 300,000 pixels. PSDs, on the other hand, are limited to 2 GB and image dimensions of 30,000 pixels. This 300,000-pixel limit is the reason why the final image has a 300,000-pixel width. I could have made the image a little bigger but I would have had to use a .kro format and I’m not sure that I would have been able to successfully blend the two images (one for the horizon and one for the city) together.

Computer: To stitch the .PSB together I used a laptop. I was worried that my laptop wouldn’t have enough horsepower to get the job done but it worked. The computer I used had the following specs: MacBook Pro (Retina, 15-inch, Mid 2015), 2.8 GHz Intel Core i7, 16GB 1600 Mhz DDR3, AMD Radeon R9 M370X 2048MB.

Hard Drive: The important thing to know when processing gigapixel images is that due to the large sizes of the images the processor speeds and RAM don’t really matter that much.

Since the processor cache and RAM fills up pretty quick when processing an image of that size the software directs everything to the hard drive where it creates something called a “page file” or “swap file” A page/swap file is a reserved portion of a hard disk that is used as an extension of random access memory (RAM) for data in RAM that hasn’t been used recently. By using a page/swap file, a computer can use more memory than what is physically installed in the computer. However, if the computer is low on drive space the computer can run slower because of the inability of the swap file to grow.

Since everything is happening on the hard disk it is really important to not only have a hard drive that is fast, but also one with a lot of space since it fills up really fast and won’t process the image if there isn’t enough space available since the swap file size can get gigantic. To process the Quito image I tried to use a fast PCI SSD that had around 500GB of space to process the image but the drive filled up. I took the computer back and got one with a 1TB PCI SSD and it was able to process the image.

Photoshop: I had to stitch one image for the horizon and another image for the background. Once these were done I opened them up in photoshop and used the eraser tool set to a large diameter to manually tool to manually blend them together. I then flattened the image and saved it as a .PSB file.

Image Tiling: I used a program called KRPano to make a tile of the images. If I uploaded the resulting .PSB file to the internet it would take forever for it to load up so people could see it. KRPano divides up the image into layers of small tiles. Each image you see is made up of a low-resolution tile. As you zoom into the image different small image tiles are quickly loaded and displayed with allows people to quickly view and explore the image without having to load the entire image. About 174,470 tiles were created for this image.

Image Upload

Once all the image tiles were created I compressed them into a .zip file. I felt that it would be easier to upload one large file instead of over 174,000 separate small files. The image upload went fine and I manually unzipped the image inside of the Hostgator server using FileZilla. It is good to check with the hosting company to make sure that they allow files to be unzipped inside their servers.

Website

Once the image was created, tiled and uploaded I made a simple website and embedded the .html file into an iframe so It could be displayed.

You can view the photo through an interactive viewer on Quito Gigapixel.

Closing

I hope that this little guide proves helpful for all of you. Gigapixel technology is really interesting and fun to try out. I have done quite a few gigapixel images but am by no means an expert and am always interested in learning more.


About the author: Jeff Cremer is a Lima, Peru-based photographer who works in the Amazon. You can find more of his work on Rainforest Expeditions and on Twitter and Instagram.



from PetaPixel https://petapixel.com/2018/02/28/created-16-gigapixel-photo-quito-ecuador/

No comments:

Post a Comment