In [1]:

from IPython.display import Image
Image(filename='Figs_ipandcg_2014/2014_Session_3.png')

Out[1]:

Python Imaging Library (PIL)

The Python Imaging Library (PIL) provide us with additional 2D image processing algorithms. PIL consists of several modules. To become acquainted with its capabilities go throuout tutorial; additional exercises you can find here (subsections 2010 Sessions 1-3).

Import functionality of PIL as Image.

In [4]:

import Image

Load image from file into variable 'im'.

In [15]:

im = Image.open('images_2014/FluorescentCells.jpg')

If you receive error message:

AttributeError: 'NoneType' object has no attribute 'bands'

you have to load data manually. Add this line:

im.load()

In [16]:

im.load()

Out[16]:

<PixelAccess at 0x6647590>

In [3]:

print im.format, im.size, im.mode

JPEG (512, 512) RGB

Split RGB imiage into 3 separate bands.

In [17]:

r,g,b = im.split()

Based on 2010 IPandCG tutorial - Part 1 and PIL handbook do following operations:

split image im and display its componentes (R,G,B). Pay attention to brightness level of each component
create new image with changed component order (e.g. B,G,R)
convert image form RGB to grayscale (L)
convert Image objec to Numpy array
rotate image
blend image
crop some Region Of Interest (ROI)

Perform various ﬁltration types of the im variable. Apply filtration for several parameter values for each filtration. Note the results for values bigger than 1 and smaller than 1. Read additional information about image filtering and edge detection (slides by prof. Paweł Strumiłło and prof. Michał Strzelecki).

ImageEnhance Module (see 2010 IPandCG tutorial - Part 2)

Sharpness()
Brightness()
Contrast()
Color()

ImageFilter Module

MinFilter()
MedianFilter()
MaxFilter()
BLUR
CONTOUR
DETAIL
EDGE_ENHANCE
FIND_EDGES
SMOOTH
SHARPEN
Kernel() for laplace filter: kernel=(1,-2,1,-2,5,-2,1,-2,1)

Compese your own averaging filter based on slides. Use 3 x 3 mask. Propose two versions of filter:

with the use of two for loops
based on matrix operation

Multi-dimensional image processing (scipy.ndimage)

Load image "blood1.bmp". Perform image thresholding e.g. with thresh equal to 126. This is not optimal threshold value. Try to find better one. Theoretical background of thresholding you can find here (by prof. Paweł Strumiłło, prof. Michał Strzelecki).

Browse capabilities of morphological opearation of ndimage library. Theory on mathematical morphology is provided in lecture. On thresholded, binary image perform some operations:

dilation
erosion
fill holes
opening
closing
morphological gradient

Write your own function that perform

binary dilation
binary erosion
binary opening
binary closing
morphological gradient

Compare with 2010 IPandCG tutorial - Part 3.

Python Imaging Library (PIL)

Linear filtering

Multi-dimensional image processing (scipy.ndimage)