A new 2D-histogram called 2D-WLDH is proposed. At the same time, a new image thresholding method based on 2D-WLDH and maximum between-cluster variance is proposed. Moreover, the corresponding fast recursive algorithm is deduced. Regional division of the proposed 2D-WLDH can avoid the shortcomings of the traditional 2D histogram. The probability of the target and background of the image can be accurately estimated by calculating the small normalized Weber Local Descriptor (WLD) value. The experimental results show that, compared with the existing corresponding algorithm, the proposed fast recursive algorithm for maximum between-cluster variance threshold selection based on 2D-WLDH, achieves better segmentation quality, which obtains uniform regions, accurate borders and robust noise resistances. Furthermore, the running time of the proposed algorithm reduces by about 84.93%.