Entropy calculation with using prediction error fields that obtain from 2-D lattice filter structure

Today 2-D signal processsing is still a current research area due to its good features: fast, efficient, easy to apply, low-cost. 2-D lattice filters are effective for modeling 2-D random fields. 2-D lattice filter structures, that use in this project, are an easy method to obtain all possible causal and noncausal 2-D models concurrently. In this method, a prediction support region is introduced and order incrementation process is carried out by prediction error fields. The method is for AR modeling of random fields. In this study, entropy has been calculated by using backward prediction error fields that have been obtained from 2-D lattice filter method and entropy is defined as an information loss between input vector and backward prediction error vector. In the thesis, three different transfer matrices have been used and eight different trials have been made for every single transfer matrix. Then entropy has been calculated for both input and backward prediction error vector. In addition, the best order incrementation way has been tried to find. For this, the difference between entropies has been investigated. The thesis consists of five parts. In the first part, makes a general preface of the topic. In the second part, is refered to usage and application areas of entropy. In the third part, is introduced to the 1-D and 2-D lattice filter structures. Forth part is comprised of entropy calculation and applications of numerical entropy calculations. In the last part is mentioned inferences, conclusions and foresights for future works.