Pore Scale invariant feature transform

PSIFT is a special SIFT methon to feature.Is is different from SIFT,it just detector the maximun value in gaussian pyramid. You can get more information form (https://www.sciencedirect.com/science/article/pii/S003132031400394X)

And i am not the auther of that paper.I just implement his methon using pyhton.By the way,this methon i written may not achiveve his desired effect.

How to use it?

There are one example test_PSIFT.py.At first ,you need twe picture which is used to be detectored in the methon.

        im1 = np.asarray(Image.open(file+'im1.png').convert('L'))
        im2 = np.asarray(Image.open(file+'im2.png').convert('L'))

im1 and im2 are images you gived.

If you want to get PSIFT feature,just read this 4 step:

• declar class 'Psift' :psift = Psift(im,octave_num=2,sigma1=1.1,upSampling=False)

• Create gaussian pyramid:psift.CreatePyramid()

• get scole poin:point_list = psift.ScolePoint(rmedge=False,curvature=10,specofic_hd=0)

• get feature by using scole point: feature, new_point = psift.GetFeature(point_list)

Application

You can match two feature picture with PSIFT:

More:https://ieeexplore.ieee.org/abstract/document/7059198/