While it is possible to correct geometric distortions after the image is captured, it is difficult to do the necessary modelling accurately and reliably. Instead, we suggest the grabbed images should simply be cropped, since the most severe distortion affects the perimeter of each video image.
To determine the area to crop, set up the Vidicon as it would be set to image a painting and capture an image of a rectangular grid -- a piece of graph paper works well as a target. Before capturing the grid, check the current video crop settings in the Preferences window and ensure that the crop is set at maximum: left 0, top 0, width 768, height 576 (these are the dimensions for a PAL signal, they may be different on your system). Also set the Aspect ratio line to 1.
Create a new video object and look for the largest rectangle with little distortion (no more than a few pixels). If you create a region on the video image (by holding down the Ctrl key and dragging down and right with the left mouse button, see §4.2) you can compare the straight edges of the region against the distorted lines of the grid. You can then expand and contract the region until you decide on the optimum area. The settings you need for the crop box can be taken from the values contained within the region object, which can be viewed by left-clicking several times on the down arrow to the left of the region object name.
Finally, you can set an aspect ratio: nip2 will automatically stretch video frames vertically by this factor. You can measure the aspect ratio of your capture card by taking a picture of something you know to be square and dividing the width in pixels by the height in pixels.
Move back to the Preferences window and enter the crop values in the General video capture section. Next time you create a new video object, you should find that it is cropped to the appropriate area. The settings you enter in the Preferences / window will be saved and automatically loaded again next time you start nip2. See appendix A.
Choosing the usable area of the image is a matter of compromise -- the smaller the area, the more images are required to build a mosaic of a particular painting. If the area chosen is too large then the amount of distortion can cause serious errors in the final mosaic.
It's possible to use the VIPS rubber sheet plug in to detect and correct geometric distortion in your images automatically. This lets you use the full area of the sensor. It is a bit fiddly, but see the rubber sheet documentation if you are determined.