今年暑假那会投了一篇icsp的会议论文，是我的水文，也有实验室同学帮我修改，由于暑假生病后基本上就是实验室的同志帮我修改后就交了。真的非常感谢他哇。论文讲的关于利用MSER和GMM联合求解星空图像中的星点质心，这片论文也与我的毕业项目相关吧。
论文的代码在www.github.com/NanYoMy上，MSER是从vl_feat上直接修改的，看他的MSER代码有点问题，在建立MSER树的时候，关于树的高度的问题，我在github上给他的pull request。
hi sir:
when i debug the mser.c. i found the value of "n_hgt" and "hgt" is not always correct. then i found the code in "mser.c":
vl_mser_pix nr_val = 0 ;
vl_uint nr_idx = 0 ;
int hgt = r [ r_idx] .height ;
int n_hgt = r [nr_idx] .height ;
r_idx = climb(r, idx) ;
nr_idx = climb(r, n_idx) ;
it means that the "n_hgt" is always the height of r[0]. because the "nr_idx" is "0" when the computer runs the code "int n_hgt = r [nr_idx] .height ;".!!.
so i modify it like this:
vl_mser_pix nr_val = 0 ;
vl_uint nr_idx = 0 ;
r_idx = climb(r, idx) ;
nr_idx = climb(r, n_idx) ;
int hgt = r [ r_idx] .height ;
int n_hgt = r [nr_idx] .height ;
it works correctly in my project. and thanks to the vl_feat, my project is success now!!!!
论文里面还有些杂乱的matlab代码，有兴趣email我