How to analyse bitmap image in python, using PIL?

Question

I was wondering how you use Python Imaging Library to analyse a simple bitmap image (say the bitmap has a thick black line at the top) to tell the program were the top of the image is. Maybe output a message when the black line is found.

Any example code would be a great help.

Answer 1

You could convert the picture to rgb which is (red,blue,green). For instance, get a picture from here:

https://github.com/panditarevolution/PIL_Play/blob/master/blackline.jpg

import PIL

# The conversion should work equally with a bitmap
img = PIL.Image.open("blackline.jpg")
rgb_im = img.convert('RGB')

rgb_im.size

This returns the size in number of pixels: (680,646). You can query the color of individual pixels with rgb_im.getpixel((x,y)) where x goes horizontal and y goes vertical, from top to bottom I believe.

So to check whether the first line is all black (or mostly black), you could do something like this:

# Get the first row rgb values
first_row = [rgb_im.getpixel((i,0)) for i in range(rgb_im.size[0])]
# Count how many pixels are black. Note that jpg is not the cleanest of all file formats. 
# Hence converting to and from jpg usually comes with some losses, i.e. changes in pixel values. 
first_row.count((0,0,0)) # --> 628
len(first_row) #--> 680

628/680 = 92% of the pixels in the first row are black.

Let's check all occurring colors in the first row with set(first_row) which gives me:

{(0, 0, 0),
 (0, 0, 2),
 (0, 1, 0),
 (1, 0, 0),
 (1, 1, 1),
 (2, 2, 0),
 (2, 2, 2),
 (4, 4, 2),
 (4, 4, 4),
 (5, 5, 3),
 (5, 7, 6),
 (6, 6, 4),
 (7, 7, 5),
 (14, 14, 12),
 (14, 14, 14),
 (35, 36, 31),
 (52, 53, 48),
 (53, 54, 46),
 (63, 64, 59),
 (64, 65, 60),
 (66, 67, 61),
 (68, 69, 61),
 (76, 77, 71),
 (79, 82, 65),
 (94, 96, 83),
 (96, 98, 87),
 (99, 101, 90),
 (101, 103, 92)}

So even if there are around 8% non black pixels, we can see that most of these are pretty monochrome, i.e. shades of gray; the rgb values are fairly close to each other for each color.

There is a good tutorial on PIL here: http://effbot.org/imagingbook/

A basic overview can be found here: http://infohost.nmt.edu/tcc/help/pubs/pil.pdf

As a bonus, and without knowing whether it's good or not (or whether it covers PIL), there is a free draft of "Programming Computer Vision with Python" available here: http://programmingcomputervision.com/