I am not sure what you mean. Are you asking about color balancing? The mosaic process uses the map information in the files to place the images into the correct map locations. If the images are in different projections or pixel sizes, then ENVI will have to reproject the file(s) to the 'base layer' that you specify. This is just warping and resampling which you can read about in the ENVI header. Color balancing uses the mean and standard deviation of all the overlapping pixels to calculate a gain and offset value that is applied to each item you want to adjust.

Thanks MariM. Yes indeed it is the Colour balancing part I am interested in. It's starting to become clearer to me thanks to your message, but can you confirm the following? Statistics (mean and SD) are created for each image, either using data from the overlapping areas or for each entire image, and a linear regression used to derive the gain and offsets between these statistics from the two images. All images are normailsed to the base layer using these gain and offsets. I'm guessing the statistics are calculated at the image sample (column) level, and hence for each unique view zenith angle. Another method I have read about uses a k-means classification to derive spectral classes first, and the statistics are then calculated for each class at the image sample (column) level. Gain and offsets are calculated in the same way as above.

Yes, thisis essentially what it is doing. The gain and offset are applied to keep all the pixel values of a given image the "same" and when you are done, all the data contain the same mean and standard deviation. It is a good way to make otherwise unrelated data "look" "the same". it is a fairly simple method.