|
|
|
nite length, therefore having a frequency re- finite impulse response filter autoregressive . infinite symmetric exponential filter (ISEF) one type of optimal edge detecter. In two dimensions, the filter is f (x, y) = a × e −p(|x|+|y|) The resulting edge detector has excellent lo- infinite-dimensional dynamical system x 0 (t) = Ax(t) + Bu(t) where x(t) is the state vector that belongs to infinite-dimensional Banach space X, u(t) is the input vector that belongs to infinite- U, A is a linear generally unbounded operator that is a gen- S(t) : X → X, for t > 0, B : u → x is a linear bounded oper- x(t, x(0), u) = S(t)x(0) + Z t 0 S(t − s)Bu(s)ds information a mathematical model of the amount of surprise contained in a message. x k is I k = − log 2 (p k ) where p k is the probability of the symbol x k . The expected value of the information of the symbols is the first order entropy of information gain for an attribute in a set of objects to be classified, a measure of the G i of the ith at- tribute A i of a set of n objects S in the clas- sification is defined as G i = I (S) − E i where I (S) is the expected information (or entropy) for the classification and E i is the expected information required for the value A i to be known. I (S) is defined as I (S) = − N c X c=1 n c n log 2 n c n where N c is the total number of classes in the classification, and n c is the number of objects in the cth class C c . E i is defined as E i = N i X k=1 n ik n I (S ik ) where S ik is the subset of S in which A i of all objects takes its kth value, N i is the number of values A i can take, n ik is the number of objects in S ik , and the information required in S ik is I (S ik ) = − N c X c=1 n ikc n ik log 2 n ikc n ik where n ikc is the number of objects in S ik belonging to class C c . information hiding a program design principle that makes available to a function information theory theory relating the information content of a message to its repre- c 2000 by CRC Press LLC |