CLASSIFICATION OF SETS OF MIXED PIXELS WITH THE HYPOTHESIS-TESTING HOUGH TRANSFORM

A new method is presented for mixed-pixel classification where the cla ssification of groups of mixed pixels is achieved by using the hypothe sis-testing Hough transform. The motivation of the work is that some o ther estimation methods based on robust statistics, such as the standa rd Hough transform, have been criticised that, although they can cope with the presence of outliers, they give poor performance in the absen ce of outliers in comparison to the least-squares-error method. The me thod proposed in the paper is demonstrated using simulated data and pr oved to perform equally well in the presence and in the absence of out liers. It is also applied to real Landsat TM data.