Images in Manifold may be projected as well as drawings. Images may be projected on the fly within maps or they may be re-projected within their native coordinate systems. Images are also re-projected automatically when they are georegistered to match a reference map, drawing or image.
Images have an implied coordinate system given by the row and column arrangement of their pixels. If an image is imported from a geographically aware image format such as GeoTIFF, it is automatically georeferenced to the correct location on Earth in geographic coordinates.
If an image is imported into Manifold form a geographically mute format such as .bmp or .jpeg, it is projected using Orthographic projection and placed with its lower left hand corner at the intersection of the Prime Meridian and the Equator (0 latitude and 0 longitude). This default georegistration decision together with size information about each pixel makes a specific choice as to where that image is located, what size it is and thus where the "center" of each pixel is located.
Just like a drawing, the coordinate properties currently in use for an image may always be seen in the Edit - Assign Projection dialog for that image. Changing any settings in this dialog does not change any pixels within the drawing - it simply changes how Manifold interprets the pixel data that already exist. To change an image's projection, we need to use the Edit - Change Projection dialog.
Although projection in images and drawings is conceptually similar, we tend to work differently with images and projections than we do with drawings and projections.
Images used for visual effects are often re-projected to achieve greater speed in a map view. Images usually contain many more pixels than drawings contain objects. The cost of re-projecting an image on the fly in a map window is therefore often much higher than the cost of re-projecting a drawing on the fly.
Images are usually highly localized views shot from directly overhead or very nearly so. They thus may be treated as already being in Orthographic projection. In contrast, most drawings or imported maps are in Latitude / Longitude coordinates and are not in any specific geographic projection.
For geographic use, drawings are most frequently imported from formats that provide a known geographic location. Images are most frequently imported from formats that provide no geographic context. Images therefore will usually need to be georeferenced.
Images are often slightly off Orthographic projection, or they are shot at angles that result in a Tilted Perspective or Space Oblique projected view in the image; however, it is very rare that we have actual numeric parameters corresponding to a particular image. Images will therefore often need to be adjusted through irregular geometric or numeric methods to fit a known good map. Drawings, in contrast, usually are transformed only through specific formulae from one known projection into another known projection.
Any re-projection of an image will add or subtract pixels whereas re-projecting a drawing will neither add nor subtract objects. For this reason, images that present raster data where the value of each pixel represents important data will not usually be re-projected.
Georegistration
Georegistration is the process of moving, re-sizing and re-projecting an image to match the geographic region it is supposed to represent. This is done using control points to match features in the image to locations in some reference map, drawing or image that is already in correct geographic context. This isn't necessary if we receive an image in a geographically aware format that automatically results in import into correct georegistration.
If we intend to use the image within a particular map it is wise to georegister the image using that map or using a drawing that employs the same projection used in that map. Doing so will georegister the image using same projection used by the map. This will eliminate the need to re-project the image "on the fly" when it is used within the map and will improve map display performance.
Once an image is georegistered to a known good geographic location and projection, we can use it like any other geographically accurate component.
Visual Images and Raster Data Images
Image formats are often used to deliver raster data such as terrain elevation or remote sensing information where the specific arrangement and value of pixels is very important. Any changes to such data sets (such as by interpolation required to change their shape into a desired projection) will add and subtract data not in the original data set.
Other types of image data are basically photographic images intended for human perception as visual images. In such cases, how the image looks to the human eye is more important than whether individual pixels are added or subtracted or otherwise changed. In many cases an overall change to the data such as improving contrast will improve the visual perception of the image.
Re-projecting an image invariably "warps" it in a way that adds or subtracts pixels. For visual images this is no problem. However, for images that represent specific raster data it may be unacceptable to change the data even if a very clever interpolation is used.
Within Manifold, we will therefore often re-project visual images without a second thought while taking care not to re-project certain types of raster data. Re-projecting photographic images into the same projection as a map view will allow the map view to operate rapidly. In the case of raster data set images we can always rely on the map view to re-project the image on the fly should it be necessary to show the raster data in a projected view. This will be slower but we will know the data has been unchanged.
To preserve the data accuracy of raster data images, they must be georegistered manually by altering their coordinate properties and not by re-projection through the georegistration process. To preserve unmodified data accuracy, raster data images must already be in some "natural" geographic projection such as Latitude / Longitude or be so close to a projection that no re-projection is necessary. Manual georegistration of such images requires opening the Edit - Assign Projection dialog and altering parameters there so the image is correctly interpreted.
Why Re-projecting an Image Adds or Deletes Pixels
All images imported into Manifold are rectangular because all raster image formats by definition have pixels arranged in rows and columns. Image windows show images as rectangles in the "native" coordinates implied by the row and column arrangement of pixels.
For example, images shot from oblique space vehicle trajectories and published as "north up" images will be shown in a rectangular format, usually with black pixels padding the extra regions needed to make the image a perfect rectangle. The image above is a Landsat 7 shot of the region near San Francisco Bay.
Although images are received as rectangles, re-projecting an image invariably changes its shape and leads to interpolation that changes the data. Re-projecting an image is, in effect, creating a new image.
Consider an image in its original rectangular format. All pixels are in locations adjacent to each other.
If we now re-project the image into some new projection (such as tilted perspective) we can imagine that the implied pixel locations will be pushed together in some regions and pulled apart in others. When shown in a map window within some projection the image will thus no longer be rectangular in shape. We should keep in mind this is only a thought experiment since images always consist of pixels adjacent to each other. Although we can imagine that pixel locations are pushed together or pulled apart in real life an image always consists of a rectangular array of pixels.
When re-projecting an image, Manifold must take the implied locations of the pixels in the new form and transfer them to a new rectangular array of pixels. The color for each pixel in the new array is determined by sampling those color values at the implied projected points within it.
When georegistering or projecting images the Size parameters specify what size the new image is supposed to be. Using fewer pixels in height and width will result in re-sampling that averages down the image. Using a higher number of pixels in height and width will provide finer sampling. By default, Manifold tries to guess at a reasonable Size setting that will preserve the approximate overall size of the image when re-projection does not result in radically different shapes.
Note that increasing the number of pixels in height and width of the re-projected image can not increase detail or improve resolution beyond that in the original image. Increasing the number of pixels in height and width beyond the size of the original image simply re-samples the same "big" original pixel over and over within adjacent new pixels.
The result of re-projection is an image that has visible pixels in the appropriate locations and otherwise has invisible pixels. Using invisible pixels is more flexible than always using black pixels to "pad" the uneven edges of images to save them in rectangles. If desired, the Paint Bucket tool can always be used to pour black color into the invisible pixels.
Note that unlike re-projecting images, re-projecting vector drawings does not involve additions to or deletions from the data set. Even though re-projecting a vector drawing changes the shapes of objects it does not add or subtract objects from the drawing nor does it add or subtract from the number of coordinate pairs used to draw each object. The shapes of all objects in drawings are defined by the coordinates used to draw those objects. Re-projecting a drawing simply changes the coordinates into different numbers but it neither adds to nor subtracts from the coordinate pairs used to draw objects.
Note: The images above are simulations and are not actual screen shots because Manifold shows images in projected views too smoothly for a real screen shot to show the effects above.
Re-projecting for Speed in Maps
When an image is displayed in a map view using a projection other than the native image projection, the map view will automatically interpolate on the fly to add and delete pixels as necessary. This can be a time-consuming process for large images and will slow down map view in such cases. Note that even though the map window is showing the image in a new projection it is not changing the actual image data. It is simply re-computing on the fly how the image would appear in the new projection.
If we get tired of a slow map window when large images must be re-projected we can open the image in an image window and permanently re-project it to the new projection using the Edit - Change Projection dialog. When the native projection used by an image is changed the image will be re-computed and pixels will be added or subtracted as necessary for the new shape. Manifold will "pad" the new, non-rectangular shape of the image with invisible pixels to maintain a rectangular format. This will speed up map view if the new projection is exactly the same as that used in the map. This is a permanent change in the actual image data.
Mystery Projections
We will often encounter images that are close to, but not quite, a direct overhead view of a geographic scene where we do not know the exact circumstances under which the image was created. Such images are effectively in a Tilted Perspective projection where we don't know the parameters - a mystery projection, as it were.
We can deal with this situation by using georegistration to warp the image with control points to make it fit into some known good drawing. Instead of using ellipsoidal trigonometric formulae to achieve a perfect mathematical transformation between two projection coordinate systems as is done with projections in drawings, this method simply uses "best fit" algorithms to distort the image so that it is a good fit as directed by the control points.
Tech Tip
When an image is used in a map, for speed of operation the map usually is specified to have the same projection as the image. This is automatically accomplished when creating the map by using the image as the first component in the map. Sometimes the map is created first, an image is added to the map and then we would like to re-project the image so it has the same projection as the map. This is easily accomplished by right clicking on the image's layer tab and choosing Project to Map. This will re-project the image to use the same projection as the map.
For advanced users: Expert users might notice that sometimes doing a Project to Map will result in an image that looks slightly different when seen in its own component window than it appears when seen in a map. The reason for this is that Project to Map tries to preserve the original resolution of the image by modifying the values of the local scale parameters, in some cases choosing values for X and Y that are different for a better match to the original resolution. (To be exactly precise, Project to Map also modifies the values of local offsets as well, although this will not change the appearance).
As a result of the modifications in local scale parameters there can occur situations where an image component window and a map window showing the same image will not look exactly the same after re-projecting an image component to a map using the Project to Map command.
To make both windows look exactly the same, follow this procedure:
1. Open the image window and choose Edit - Change Projection.
2. Select the coordinate system to be the same as the map.
3. Press the Suggest button to let the system compute optimal values for local scales and offsets. These are the same values that would be used by the Project to Map command, were it to be used.
4. Modify the values of the local scale parameters so that they are equal to each other. There is no need to modify the values of local offset parameters.
5. Press OK to re-project the image.
See Also
Changing a Component's Projection