http://en.wikipedia.org/wiki/Slit-scan_photography

sscan.py

#!/usr/bin/env python

# ./sscan.py input.mov rownumber output.png

# This is meant to be hyper-simple and makes
# some assumptions like: you want a row (not
# a column), the video is RGB (not gray), etc.

# Bug: frame_count is sometimes fractional.
# int() and the "if not okay" are workarounds.

import numpy as np
import cv2
from sys import argv

movf, row, outf = argv[1:]
row = int(row)

slice_width=12

viddy = cv2.VideoCapture(movf)
width = int(viddy.get(3)) # can you tell this API is by C programmers?
height = int(viddy.get(4))
frame_count = int(viddy.get(7))

print width, height, frame_count
rows = np.empty((height, frame_count*slice_width, 3)) # 3 channels

for frame_n in range(frame_count):
    print("%s/%s" % (frame_n, frame_count)) # or not
    okay, frame = viddy.read()
    if not okay: break
    for i in range(0, slice_width):
        rows[:,(slice_width*frame_n)+i] = frame[:,row+i]

cv2.imwrite(outf, rows)