The adjacency mask set to 1 when Si and Sj share a common boundary. How to check if these two share a common boundary?

[luciaL]

Thank you so much for your code.
I don't understand this line: sub = self.segments[i:i + 2, j:j + 2]
Could you please explain how to judge if the two superpixels share a common boundary?
`
def get_A(self, sigma: float):
'''
根据 segments 判定邻接矩阵
:return:
'''
A = np.zeros([self.superpixel_count, self.superpixel_count], dtype=np.float32)
(h, w) = self.segments.shape
for i in range(h - 2):
for j in range(w - 2):
sub = self.segments[i:i + 2, j:j + 2]
sub_max = np.max(sub).astype(np.int32)
sub_min = np.min(sub).astype(np.int32)
# if len(sub_set)>1:
if sub_max != sub_min:
idx1 = sub_max
idx2 = sub_min
if A[idx1, idx2] != 0:
continue

                pix1 = self.S[idx1]
                pix2 = self.S[idx2]
                diss = np.exp(-np.sum(np.square(pix1 - pix2)) / sigma ** 2)
                A[idx1, idx2] = A[idx2, idx1] = diss 

`

[qichaoliu]

Thanks. In fact, I adopted a very simple suboptimal solution. I used a very small window （2x2 or 3x3）to detect the boundaries, i.e., if the classes of pixels in this window exceed 2, then it must be on the boundary. Just like the edge detector.

Best regards.