Different image output with same inversed transformation

Question

I got different image output while transformation stayed the same

im=np.ones((128,128))*255
src = np.array([
    [56., 60.],
    [81., 60.],
    [64., 80.],
    [49., 100.],
    [78., 100.]], dtype=np.float32)
q=np.array([
    [58.166317,39.124466],
    [86.10652,26.917824],
    [106.29474,39.68392],
    [90.64203,72.82854],
    [115.17699,63.91104]],dtype=np.float32)
tt=trans.SimilarityTransform()
tt.estimate(q,src)
dst1 = trans.warp(im,tt.inverse,preserve_range=True,output_shape=(128,128))
tt.estimate(src,q)
dst = trans.warp(im,tt,preserve_range=True,output_shape=(128,128))

So,you can see the difference, here are outputs:

Why are not they the same?

UPD: Main issue is in "estimate" function, that estimate(q,src)!=(estimate(src,q))^-1. Still open question: why

Answer 1

To clarify the question: affine transformations is a group of transformations t:R2->R2, where you have one determined inverse element for each (in our case - inverse transformation). In my question it appears that we have different transformations to the same source points (two inverse elements). Actually, there is no mistake anywhere, just misunderstanding .inverse and .estimate function. t.estimate(q, src).inverse !=t.estimate(src, q) but t.estimate(q, src).inverse ==t.estimate(t(q),q)