How to iterate over columns of an image?

I want to transform images like this to add an effect to pictures in django as described here.

I decided to implement it as a process for the great django-imagekit/photologue

My knowledge of PIL isn't very good, so my question is

How to intend a column of pixels by a sinus-offset in PIL?

any hints (code, lins, general thoughts) are welcomed

Here is an quick-n-dirty example which should bring you to the right direction:

from PIL import Image, ImageOps
import math

src = Image.open('arched.jpg')

ampl = step = 10

img = ImageOps.expand(src, border=ampl*4, fill='white')
size = img.size

straight_mesh = {}
distorted_mesh = {}
for y in range(size[1]//step-1):
    py = step*(y+1)
    dx = -int(round(ampl*math.sin(py*math.pi/size[1])))
    print dx 
    for x in range(size[0]//step-1):
        px = step*(x+1)
        straight_mesh[x, y] = (px, py)
        distorted_mesh[x, y] = (px+dx, py)
transform = []
for x in range(size[0]//step-2):
    for y in range(size[1]//step-2):
        transform.append((
            map(int, straight_mesh[x, y] + straight_mesh[x+1, y+1]),
            map(int, distorted_mesh[x, y] + distorted_mesh[x, y+1] + \
                    distorted_mesh[x+1, y+1] + distorted_mesh[x+1, y])
        ))
img = img.transform(size, Image.MESH, transform, Image.BICUBIC)
img = ImageOps.crop(img, border=ampl*2)
img.save('result.jpg')