The first technique of compression we want to consider is prediction.  In order to reduce the amount of data transmitted, we wish to find some redundancy that may be exploited, that is, some predictable property of the picture data.  Once such a property has been identified, it can be used to predict the data to be transmitted, and then we may transmit only the difference of the actual quantity and the predicted quantity.  This technique by itself does not reduce the amount of data, but can be combined with statistical entropy coding to reduce the data rate.  

Predictive entropy coding usually consists of some type of differential coding followed by a Huffman entropy code.  Let's look at a concrete example:

NEXT: Example: Coding based on clustering of pixel values