Parent Topic: Theory

Mosaicking

Mosaicking is the blending together of several arbitrarily shaped images to form one large, radiometrically balanced image, so that the boundaries between the original images are not easily seen. Any number of geocoded images can be blended together along user-specified cut lines (polygons).

Mosaicking is a special case of Geometric Correction where the registration takes place into an already existing image. If GCPs are collected, the uncorrected image is transformed according to the derived polynomial, into the georeferenced image. If no GCPs are provided, but both images already have compatible georeferencing, then an appropriate translation and scaling will be applied instead of a polynomial transformation.

Mosaic Cut Line

When images to be mosaicked overlap, it is useful for the user to specify a cut line determining which image takes precedence in different areas. This is done in GCPWorks on the ``Mosaic Area Collection'' panel.

The cut line is a polygon, enclosing the region to be replaced in the georeferenced image. Any of the image pixels within the polygon will be replaced by pixels from the uncorrected image if possible. Georeferenced pixels will not be replaced if the corresponding location is off the uncorrected image, or if the pixel value on the uncorrected image is the ``Background Value''.

The ``Mosaic Area Collection'' panel also contains options for loading and saving the cut line to or from a vector segment on a PCIDSK database.

Note that it is currently only possible to have a single mosaic cut line per mosaic operation; however, it is possible to perform many mosaic operations from one uncorrected image to one georeferenced image, each time selecting a region with a different cut line. To do this, it is necessary to alternate between the ``Mosaic Area Collection'' panel and the ``Disk to Disk Registration'' panel to collect cut lines, and perform mosaics.

Blending

Feathering (blending) can be performed along seams between the georeferenced and uncorrected images, making the seams unnoticeable. Cut lines can be hand-drawn or imported from a vector segment.

Blending makes the sharp changes occurring at the cut line appear more gradual by altering pixel values at the cut line. The blend width sets the number of pixels over which the blending will take place. For example, if the blend width is 8, the pixel values 4 pixels into the uncorrected image from the cut line will be composed 100% of the original uncorrected image pixel value. Similarly, the pixel values 4 pixels into the georeferenced image from the cut line will be composed 100% of the original georeferenced image pixel value. At the cut line, the pixel values will be composed 50% of the uncorrected image pixel values and 50% of the georeferenced image pixel values.

It is important to note that blending will not fix poor registration problems. For instance, if a road does not match up well at a cut line, the match up will be no better and may even be more obvious after blending is complete.

The blending width may be altered on the ``Disk to Disk Registration'' and ``Pre-Registration Checking'' panels.

Colour Balancing

Uncorrected images can be altered to make the mosaic less noticeable. Colour balancing is achieved by calculating the lookup table required to approximately match the uncorrected image to the georeferenced image. The creation of this new lookup table is called ``histogram matching''. An image histogram is a graph showing the number of pixels with a given brightness, for each possible brightness value. In histogram matching, the image histograms of the uncorrected and georeferenced images are compared, and an attempt is made to make the uncorrected image histogram match the georeferenced image histogram by creating a ``histogram matching lookup table''. All values in the uncorrected image will be passed through this lookup table to create the ``histogram equalized'' (colour balanced) image.

The ``Colour Matching'' panel in GCPWorks provides the opportunity to select areas for histogram comparison. It is important that appropriate selections are made. Like areas should be compared. Comparing bright fields to dark mountains will produce a histogram equalized image with dark fields and very dark mountains, or light mountains and very bright fields (depending on which is the georeferenced and which is the uncorrected).