merge

    diff = np.empty((row,col))

    diff[0,:] = np.zeros(col)       # keep the first row from the image

    diff[:,0] = np.zeros(row)

    predict = np.empty([row,col])     # create a empty matrix to update prediction

    predict[0,:] = image[0,:]       # keep the first row from the image

    predict[:,0] = image[:,0]       # keep the first columen from the image

    predict[-1,:] = image[-1,:]       # keep the first row from the image

    predict[:,-1] = image[:,-1]       # keep the first columen from the image

    diff = np.empty([row,col])

    diff[0,:] = np.zeros(col)       # keep the first row from the image

    diff[:,0] = np.zeros(row)

    diff[-1,:] = np.zeros(col)       # keep the first row from the image

    diff[:,-1] = np.zeros(row)

    for r in range(1,row-1):                  # loop through the rth row

        for c in range(1,col-1):              # loop through the cth column

            surrounding = np.array([predict[r-1,c-1], predict[r-1,c], predict[r-1,c+1], predict[r,c-1]])

            predict[r,c] = np.mean(surrounding)       # take the mean of the previous 4 pixels

            diff[r,c] = (np.max(surrounding)-np.min(surrounding))

    predict = np.ravel(predict)

    diff = np.ravel(diff)

    n = len(predict)

    fig = plt.figure()

    ax1 = fig.add_subplot(111, projection='3d')

    z3 = np.zeros(n)

    dx = np.ones(n)

    dy = np.ones(n)

    dz = np.arange(n)

    ax1.bar3d(predict, diff, z3, dx, dy, dz, color="red")

    ax1.axis('off')

    plt.show()

    return image, predict, diff

if __name__ == '__main__':

    """For boundary cases: Start by grabbing the shape of the images and saving those

    error = np.abs(image-predict)

    plot_hist(images)

 No newline at end of file