Decision-Tree-Based Interpolation for Multidimensional Signal Compression

We investigate a decision-tree-based interpolator as part of multidimensional signal compression methods. This interpolator selects the interpolating function at each signal point through a decision tree. We propose a learning algorithm for this decision tree. This algorithm is based on a recursive procedure for calculating the entropy of quantized interpolation errors. We adapt the decision-tree-based interpolator for the hierarchical compression method. We propose a multidimensional local feature and a set of interpolating functions for this compression method. We research the decision-tree-based interpolator for compression of natural multi-dimensional signals. The experimental results prove that this interpolator can significantly (up to 30%) increase the efficiency of the hierarchical compression method.