Wavelet-Packet Content for Positive Operators
Abstract: We give a simple way to attach ``content" to the nodes of a wavelet packet tree when a positive operator is given. At a fixed packet depth, the packet projections split the operator into positive pieces, and this decomposition induces a boundary measure on the packet path space, together with vector-dependent densities that show how energy is distributed across the tree. We then study a sequential extraction procedure and two depth-fixed greedy rules for choosing packet blocks, one based on trace weights and one based on Hilbert-Schmidt weights. The main results are explicit geometric decay estimates for the remainder under these greedy removals. In the Hilbert-Schmidt case we also isolate a coherence quantity that measures how close the operator is to being block-diagonal in the packet partition. We close with a concrete patch-based denoising procedure for images, where packet blocks are selected by these content weights computed from an empirical second-moment operator; the construction ensures that both the approximants and the remainders stay positive at every step.
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Collections
Sign up for free to add this paper to one or more collections.