Resample an Image
Choose Reformat>Resample off the top menu bar.
The following screen should appear:
In the "Input file:" white box you should type in the file name of the imported scanned image or double click on the box. If you double-click then choose the image off the list by double-click on it.
In the "Output file:" white box type a name for the georeferenced file, no extension is needed.
In the "Correspondence file:" box type in the name of the .cor file that you entered the coordinate information.
In the "Reference system:" box choose "spc27ct1" for CT state plane NAD27 coordinate system for the output image. All data on MAGIC is in CT state plane NAD27 except for soils data. Therefore it is highly reccomended that unless you have another reference system in mind you use CT state plane NAD27.
If you choose CT state plane NAD27 then "Reference units:" should be set to feet. If you choose a different coordinate system then your units will correspond to whatever units are used in your reference system.
The "Mapping function:" you choose depends on the number of ground control points you have in your .cor file and the amount of distortion in your photograph. For most simple cases linear should be adequate.
"Resampling type:" determines how the new pixel values will be written to the output file. For Nearest neighbor, the pixel with the closest coordinates will be written to the file, with Bilinear an average is used. If you are working with photographs it is highly reccomended that you use Nearest neighbor. Bilinear will distort your image and will change the way your photograph looks.
The next step is very important. You must input the minimum and maximum X and Y coordinates of the final output image or map. The implication of this is that any sub-region of the original image may be extracted. Coordinates must be expressed in the new output coordinate system. If the final output is an image, you will also need to input the number of columns and rows that will span that region. You are free to set any resolution you wish here but bear in mind the original resolution of the data. In general you should set the number of columns and rows to the number in the original imported image.
Hit the OK button and a screen should appear that allows you to calculate the Root Mean Squared (RMS) Error of your transformation. RMS error gives you an indication of the varability of the new values around their true values. It is an aggregate measure taken from all the resampling points you use in the transformation. In general, the lower the RMS error the closer your output image coordinates will be to the real world (i.e. the greater the coordinate accuracy of your final product). If a pixel on your image represents about 10 inches on the ground, then a resample value of 1 means that your average error per pixel is about 10 inches.
The Idrisi help file states:
-
The RMS for raster images should be less than one half the resolution of the input image. Generally, it is recommended that the control point with the highest residual be omitted in order to reduce the RMS. It is important, however, that the correspondence file retain control points in all areas within the image.
Some times it will not be possible to get the RMS error low enough. We reccomend trying several different point combinations and produce several output files. You can compare the output files by displaying them and overlaying some vector coverage like roads or rivers. Then you can retain the file which has the best accuracy across the area you are interested in using.
When your RMS error is acceptable then you should hit the OK button. At the bottom of the main Idrisi screen you will see a progress bar in the lower left corner. For large files this process could be quite slow. After the process is done, Idrisi will display the image to show you how the georeferencing process changed the image. If the new image is too distorted you should go through this process again but retain more GCPs during the resample process. Remember that the correction will probably not be the same across the whole image. The center of your image will probably show a closer correspondence to real world coordinates than other areas of the photo. If the area you are interested in is not in the cneter then be sure to put quite a few GCPs in the your area. If the image looks good to you then you can export it to various formats. There are several options for what you can do with this final product from here.

