Bi-cubic interpolation for image scaling

Question

I'm trying to implement a very basic Bi-Cubic algo to use it for image scaling. I used other questions on stack overflow and tried to just translate the code into the Python.

It executes correctly w/o any errors, I've checked the math a couple of times but I cant find the reason of that kind of problem seems like everything should be running smooth but as a result I get this:

Any help appreciated.

Here's the source:

def getBicPixelChannel(img,x,y,channel):
if x < img.shape[1] & y < img.shape[0]:
    return img[y,x,channel]

return 0


def Bicubic(img, rate):
new_w = int(math.ceil(float(img.shape[1]) * rate))
new_h = int(math.ceil(float(img.shape[0]) * rate))

new_img = np.zeros((new_w, new_h, 3))

x_rate = float(img.shape[1]) / new_img.shape[1]
y_rate = float(img.shape[0]) / new_img.shape[0]

C = np.zeros(5)

for hi in range(new_img.shape[0]):
    for wi in range(new_img.shape[1]):

        x_int = int(wi * x_rate)
        y_int = int(hi * y_rate)

        dx = x_rate * wi - x_int
        dy = y_rate * hi - y_int

        for channel in range(new_img.shape[2]):
            for jj in range(0,4):
                o_y = y_int - 1 + jj
                a0 = getBicPixelChannel(img,x_int,o_y, channel)
                d0 = getBicPixelChannel(img,x_int - 1,o_y, channel) - a0
                d2 = getBicPixelChannel(img,x_int + 1,o_y, channel) - a0
                d3 = getBicPixelChannel(img,x_int + 2,o_y, channel) - a0

                a1 = -1./3 * d0 + d2 - 1./6 * d3
                a2 = 1./2 * d0 + 1./2 * d2
                a3 = -1./6 * d0 - 1./2 * d2 + 1./6 * d3
                C[jj] = a0 + a1 * dx + a2 * dx * dx + a3 * dx * dx * dx

            d0 = C[0] - C[1]
            d2 = C[2] - C[1]
            d3 = C[3] - C[1]
            a0 = C[1]
            a1 = -1. / 3 * d0 + d2 - 1. / 6 * d3
            a2 = 1. / 2 * d0 + 1. / 2 * d2
            a3 = -1. / 6 * d0 - 1. / 2 * d2 + 1. / 6 * d3
            new_img[hi, wi, channel] = a0 + a1 * dy + a2 * dy * dy + a3 * dy * dy * dy

return new_img

Answer 1

I think your problem is in this line of code:

if x < img.shape[1] & y < img.shape[0]:

From what I gather, & is the bit-wise AND operator in Python, and it has a higher precedence than <. Thus what you are computing is:

if x < ( img.shape[1] & y ) < img.shape[0]:

(whatever that means...)

Instead, try this:

if (x < img.shape[1]) and (y < img.shape[0]):

and is Python's boolean AND operator.