Image blur assessment has always played an important role in the image quality assessment (IQA). In fact, many methods have been used to measure the blur of an image, but we find that lots of the traditional methods can't effectively compare the blur among images with different contents, leading to the blur measuring more difficult. Recently, we find a perfect method based on the idea that it is difficult for us to find the differences between an already blurred picture and a re-blurred picture based on the former. As a result, Crete et al. use this character of human eyes to assess the blur of an image. and study it carefully. By analyzing the method, though it's good enough, we find some shortcomings of it and propose two improved algorithms based on it: a new blur metric based on the variation of the neighboring differences around edge and a blur metric based on neighboring variation of second difference matrix. The two new methods have been proved that they are all improved to some extent compared with the original algorithm by experiments. Being helpful to the blur assessment, the improved methods are better in veracity, consistency and monotonousness. What's more, they provide a novel perspective to develop blur metrics.