A fast technique for identifying zerotrees in the EZW algorithm

The embedded zerotree wavelet (EZW) algorithm has become a popular benchmark for image compression algorithms. The current paper presents a fast technique for identifying zerotrees for all dominant passes in the encoder prior to encoding. The key is initializing a data structure called a zerotree map with the largest power of two smaller than a given coefficient, and then completing the zerotree map by bitwise-ORing the values of potential parents with those of their children. This simple operation can be performed in parallel with the wavelet transform operation and thus is efficient for both hardware and software implementations