What is the difference in accuracy between a single pixel and a clump of pixels (i.e., how accurate are the data)?
The concept of accuracy in digital satellite image mapping is not as simple as it first appears. Since a pixel is representative of an area of ground, usually a square, the light reflected from the entire area is averaged to get one pixel value though it may contain many different materials. When the entire area is covered in a single material or vegetation cover, identification of the material is fairly straightforward. However, when a pixel falls on two or more materials, this causes problems for identification. These areas are called “mixed pixels.” For this reason, accuracy assessment is usually performed on homogeneous groups, or neighborhoods of pixels, or clumps. By identifying the clumps and field checking them, we can get a measure, statistical and empirical, of the accuracy and “fitness for use” of the data. This means that a given accuracy figure corresponds to the data’s ability to represent homogeneous, or spectrally pure, cover types.
