如何将张量分为NXN块并再次从这些块重建第一个？

Question

I have a tensor with shape (1,32,32) that the first index in shape shows the number of channels and my images are gray. now I need to divide this tensor into 8x8 blocks and do some processes on each block like DCT and after changes, replace these new blocks with the previous one. we have mattocell and celltomat in Matlab, but in keras, I do not know how can I do this? could you please tell your suggestions in Keras? Thank you.

I used this code:

def slicAndJpeg(img):
    for i in range (int(img.shape[1].value/8)):
        for j in range(int(img.shape[2].value/8)):
            temp=img[:,i*8:i*8+8,j*8:j*8+8]
            tempb=K.zeros((8,8))
            DCT(temp,tempb)

but I do not know it is true or not?! could you please tell me it is possible to use the above code for tensors in keras? Thanks

Answer 1

I used the before mentioned code and it works. based on my experiments we can do this by for loops, but if somebody has a better suggestion please put it here.

 def slicAndJpeg(img):
        for i in range (int(img.shape[1].value/8)):
            for j in range(int(img.shape[2].value/8)):
                temp=img[:,i*8:i*8+8,j*8:j*8+8]
                tempb=K.zeros((8,8))
                DCT(temp,tempb)