The question 'What is a mosaic image?' was posted by one of the students ("emmagrace") in the I'm a Scientist Event. Of course, I gave an answer, but this one really needed pictures to go with it.
So, as a special feature, here's what a mosaic image is, along with a complete example of how we make some of the panorama photographs that feature on the KAIRA web log. This has been written especially for the students so, for the imaging scientist experts out there, bear with me while I explain it!
Of course, as the other scientists answered on the website, a mosaic is indeed where an artist uses hundreds of tiny coloured tiles to make a picture. There are many splendid examples of mosaics both recent and from ancient times. A quick look at the Wikipedia article will reveal some spectacular examples.
In a way, any digital photograph is a mosaic of thousands, if not millions, of tiny squares called pixels. Some images, known as photographic mosaics are composed such that each pixel is, in itself, an image.
However, there is another type of mosaic image. And this one is directly relevant to many scientific applications. The term is particularly used in astronomy and microscopy, mostly as this is where the technique is needed due to limitations of the optics of telescopes and microscopes. In these cases, a mosaic is a large image, made up several parts of and image. You can do the same thing with normal photography, in fact. These images are also called mosaics, as well as 'stiched', 'composite' or 'tiled' images.
Mosaic images like this are wonderful when the field of view of the scene is far larger than the field of view of the camera. Because the receiver array at KAIRA is so big, and the scenery so wide, we often used mosaicing to create our panorama images.
Here's how it works.
Let's imagine that you have a scene that you want to photograph... something like this. It is a wide open space, with some mountains in the background, clouds and dirt. It is where the KAIRA construction project is happening and we want a photograph of it all!
The problem is that our camera has a limited field of view. It can't see everything at once. You have to choose which bit you look at. For example, maybe just the middle.
If you take this photograph sure, it looks great, but we miss out on the edges and can't see the mountains.
So, what we do instead is take 3 photographs. They need to be slightly overlapping (and we'll see why later). But for now, you could simply line them up. Not bad. You get a rough idea, but it is still not too good.
Of course, you can cut-and-paste the images a bit, but there are other problems. The images are not necessarily flat (meaning, that they are a bit distorted at the edges). Also, they may have different exposures, meaning that one image may be slightly darker than another.
Take a look at the images now. They overlap nicely, but it is still very obvious where the line between them is.
To fix this, we compare the images with a special computer programme. What we do is mark matching points in one image, with the exact matching point in the other image. It might be a peak on the mountain or a light-coloured stone, or a part of the timber. It doesn't matter. The point is that we are telling the computer that this point on this image, matches that point on the other image.
Here's what it looks like, with the matching points flagged with little numbers.
If you then run the programme, it will work out the distortion of the individual images and correct for them. Look at the red and blue outlines of the centre and right-hand image. Although you may not have noticed the distortion in the original, it is actually quite pronounced and this prevents us from getting a good match.
Additionally, the software can adjust the brightness of each image, so that they match.
Finally, the image is stitched together and blended smoothly. If you didn't already know, you may not realise that this image is not one, but three images... all neatly mosaiced together.
And that is what the KAIRA build site looked like on 14th June 2011.
As I mentioned, we have used mosaicing for other images on our web log, such as a view of Saana made from 17 images and a view of the winter testing made from 50 images. And this technique is used in astronomy and geophysics for mapping, imaging and aligning huge sets of data.
PS: You can click on any of the images to see larger versions. Try it... especially on the last one!