The goal of entropy coding is to code your data such that the length of the code associated with a symbol is close to $-\log_2 p$, where $p$ is the probability of the symbol (Shannon's limit). This can be achieved through different techniques: Huffman coding is one of them, but arithmetic coding is another option - or if the distribution of the coefficients can be known in advance, a pre-computed code can be specified. The choice of an entropy coding method is motivated by performance/complexity trade-offs, and by the known (or unknown) distribution of the coefficients. If two transforms generate the same distribution of coefficients, the same entropy coding methods can be used for both.
Since the diagram from your question hints that your application is JPEG2000 compression, I suggest you to look into the JPEG2000 specifications to see which entropy coding method they have actually specified.