On the other hand, for the histogram modification of PEE, many adaptive mechanisms have been exploited as well. In [34], for a given payload, Xuan et al. propose selecting the best threshold as well as the most suitable embedding region to minimize the em- bedding distortion. In [28], Wang et al. propose calculating the embedding distortion for each possible choice of expansion bins, and then select the optimal one that minimize the embedding distortion. In [17], multiple pairs of histogram bins are selected for expansion based on the efficiency of modifying each histogram bin pair, where the efficiency is defined as the increasing em- bedded bits for each decrease of PSNR value. In [35], Coatrieux et al. propose utilizing a classification process to employ HS in one part of image pixels, while the other part is adopted the dynamic PEH modification mechanism. In [21], instead of a single PEH, Li et al. propose generating a sequence of histograms with different complexity levels, and the expansion bins are adaptively selected for each generated histogram. In a recent work [23], Wang et al. propose formulating multiple HS as a rate-distortion optimization problem, and the proper peak and zero bin pairs are determined by employing the genetic algorithm.