Consider a situation where you are standing on top of a mountain or some other beautiful natural scenery. You are enjoying a beautiful view that seems to span from far left to far right and you want to take a nice picture of the whole thing. Your camera allows you to capture only a limited field of view. So to capture the whole scene, you will have to capture multiple images. Doesn’t feel exactly the same watching it in pieces, does it? We really want to capture the beauty within a single image. You can certainly record a video and capture the whole scene, but what if you want to print it out? This is where panoramic photography technique comes in. Panoramic images are images with elongated field of view. The image above is one such example. These images cannot be captured with a single camera click because the field of view is limited. So how do we do capture panoramic images?
What is panoramic photography?
Panoramic photography is a technique that stitches multiple images from the same camera together to form a single, wide photograph. The word “panorama” actually means “all sight” in Greek. We just capture multiple images spanning the entire scene and this technique will stitch all the images together. With the invention of personal computing and digital photography, it is now much easier to stitch digital images together using specialized software. This is quite popular among photographers who like landscapes, architecture and cityscapes.
Is it difficult to do?
Panoramic Photography can get quite complex and expensive, depending on what you are trying to do. If you are creating panoramic images in architectural photography, you will need the camera and lens to be properly calibrated to prevent curved lines, distortions and improper stitches of close objects. At the same time, you can successfully take great landscape panoramic images without investing on any camera equipment, as long as you know how to do it right. Modern technologies are focusing more towards robustness which need minimum responsibility from the user. There are apps for panoramic photography on smartphones and tablets. You can download one of them and try it out.
How do we stitch the images to form panoramic images?
Once we capture multiple images, we have to combine to correctly to get the final panoramic image. Images will have overlapping fields of view and we have to combine them appropriately. Image stitching involved three stages: registration, calibration and blending.
Image Registration: Once we have all the images, we take the first image and keep attaching the images to it one by one. Let’s consider two images. We need to know what portions of the images are overlapping so that we know how to combine then. As you can see in the image shown, a portion of the first image overlaps with a portion of the second image. Same is true with second and third, and so on. We need to know exactly where the corresponding points are, else there will be blurriness and discontinuities. We extract features from these two images and match them to obtain the required transformation. Different implementations use different features like SIFT, SURF, Corner features etc. They are pretty famous and there are a lot of tutorials available. You can read up more about them if you want.
Image Calibration: This step is more related to the camera distortions. Given a perfect set of images, you wouldn’t need to worry about this step much. Calibration actually aims at reducing distortions, aberrations, exposure differences etc caused by the camera.
Image Blending: Now that we have figured out how much each image overlaps with the next image, we have to stitch them together to get a seamless wide angle output image. There are many variables to take care of such as geometric distortions, color adjustments between multiple images, blurring, jerks, ghosting etc. The users cannot capture perfectly still images, so we need all these technologies to take care of that. It’s difficult to hold a mobile device perfectly still when we capture these images. Even though the device is still, the exposure might change between multiple images, which will cause discontinuities in the final image. Human eye is very sharp at noticing such things!
There are different types of panoramic images:
Wide angle: Anything which covers less than 180 degrees field of view. These are very common when users capture pano images.
180 degree: These images cover 180 degrees field of view. These look very wide and cover a very large area.
360 degree: These images cover 360 degrees field of view. That’s an entire circle! You hold your device up and make a complete circle. These are super wide angle images and you can fit these images on cylindrical objects (like a bottle) and it will look like a single seamless image.
Spherical: These images cover an entire sphere. It’s like capturing anything and everything in your field of view. They are then converted into square spherical images. You can wrap it around a ball and it will look like a single seamless image. People actually came up with a throwable ball camera to capture spherical panoramic images. Check out this video to see what I’m talking about.
Users just start the app and move their device around. The software clicks multiple shots and then stitches them together to give the final output. This is one of the cool applications of computer vision and photographers absolutely love this technology. Natural sceneries look really good when captured using panoramic photography.