Binary Image Proc: week 2

 Getting a binary image

 Connected components extraction
 Morphological filtering
 Extracting numerical features from
regions

Stockman CSE803 Fall 2008 1

Quick look at thresholding

 Separate objects from background.

 2 class or many class problem?
 How to do it?
 Discuss methods later.

Stockman CSE803 Fall 2008 2

 Cherry image shows 3 regions

 Background is black
 Healthy cherry is bright
 Bruise is medium dark Use this gray
Histogram shows two cherry
value to separate

regions (black background has
been removed)

Stockman CSE803 Fall 2008 3

 Choosing a threshold
 Common to find the deepest valley
between two modes of bimodal histogram
 Or, can level-slice using the intensities
values a and b that bracket the mode of
the desired objects
 Can fit two or more Gaussian curves to
the histogram
 Can do optimization on above (Ohta et al)

Stockman CSE803 Fall 2008 4

 Connected components
 Assume thresholding obtained binary image
 Aggregate pixels of each object
 2 different program controls
 Different uses of data structures
 Related to paint/search algs
 Compute features from each object region

Stockman CSE803 Fall 2008 5

Notes on Binary Image Proc
 Connected Components Algorithms
 Separate objects from background
 Aggregate pixels for each object
 Compute features for each object
 Different ways of program control
 Different uses of data structures
 Related to paint/search algs
Stockman CSE803 Fall 2008 6
Example red blood cell image
 Many blood cells are
separate objects
 Many touch – bad!
 Salt and pepper
noise from
thresholding
 How useable is this
data?

Stockman CSE803 Fall 2008 7

Cleaning up thresholding results
 Can delete object
pixels on boundary
to better separate
parts.
 Can fill small holes
 Can delete tiny
objects
 (last 2 are “salt-and-
pepper” noise)
Stockman CSE803 Fall 2008 8
Removing salt-and-pepper
 Change pixels all of
whose neighbors are
different (coercion!):
see hole filled at right
 Delete objects that
are tiny relative to
targets: see some
islands removed at
right

Stockman CSE803 Fall 2008 9

Simple morphological cleanup

 Can be done just after thresholding

-- remove salt and pepper
 Can be done after connected
components are extracted
-- discard regions that are too small
or too large to be the target

Stockman CSE803 Fall 2008 10

CC analysis of red blood cells
 63 separate
objects
detected
 Single cells
have area
about 50
 Noise spots
 Gobs of cells

Stockman CSE803 Fall 2008 11

More control of imaging
 More uniform
objects
 More uniform
background
 Thresholding
works
 Objects actually
separated

Stockman CSE803 Fall 2008 12

Results of “pacmen” analysis
 15 objects
detected
 Location known
 Area known
 3 distinct
clusters of 5
values of area;
85, 145, 293

Stockman CSE803 Fall 2008 13

Results of “coloring” objects
 Each object
is a
connected
set of pixels
 Object label
is “color”
 How is this
done?

Stockman CSE803 Fall 2008 14

Extracting components: Alg A

 Collect connected foreground pixels into

separate objects – label pixels with
same color
 A) collect by “random access” of pixels
using “paint” or “fill” algorithm

Stockman CSE803 Fall 2008 15

paint/fill algorithm
 Obj region
must be
bounded by
background
 Start at any
pixel [r,c]
inside obj
 Recursively
color
neighbors
Stockman CSE803 Fall 2008 16
Events of paint/fill algorithm
 PP denotes “processing point”
 If PP outside image, return to prior PP
 If PP already labeled, return to prior PP
 If PP is backgr. pixel, return to prior PP
 If PP is unlabeled obj pixel, then
1) label PP with current color code
2) recursively label neighbors N1, …, N8
(or N1, …, N4)

Stockman CSE803 Fall 2008 17

 Recursive Paint/Fill Alg: 1 region

 Color closed FILL ( I, r, c, L)

boundary with L If [r,c] is out, return
 Choose pixel [r,c] If I[r,c] == L, return
inside boundary I[r,c]  L // color it
 Call FILL For all neighbors [rn,cn]
FILL(I, rn, cn, L)

Stockman CSE803 Fall 2008 18

Connected components using
recursive Paint/Fill
 Raster scan until object pixel found
 Assign new color for new object
 Search through all connected neighbors
until the entire object is labeled
 Return to the raster scan to search for
another object pixel

Stockman CSE803 Fall 2008 19

Extracting 5 objects

Stockman CSE803 Fall 2008 20

Outside material to cover
 Look at C++ functions for raster
scanning and pixel “propagation”
 Study related fill algorithm
 Discuss how the recursion works
 Prove that all pixels connected to the
start pixel must be colored the same

Stockman CSE803 Fall 2008 21

Alg B: raster scan control

Visit each image pixel once, going by

row and then column.
Propagate color to neighbors below
and to the right.
Keep track of merging colors.
Stockman CSE803 Fall 2008 22
Raster scanning control

Stockman CSE803 Fall 2008 23

 Events controlled by neighbors
 If all Ni background,
then PP gets new
color code
 If all Ni same color
L, then PP gets L
 If Ni != Nj, then
take smallest code
and “make” all
same
 See Ch 3.4 of S&S
Stockman CSE803 Fall 2008 24
 Merging connecting regions

Detect and record

merges while
raster scanning.
Use equivalence
table to recode

Stockman CSE803 Fall 2008 25

 alg A versus alg B
 Visits pixels more than
once  “visits” each pixel once
 Needs full image  Needs only 2 rows of
 Recursion or stacking image at a time
slower than B  Need to merge colors
 No need to recolor and region features
when regions merge
 Can compute features on  Typically faster
the fly
 Can quit if search object
 Not suited for heuristic
found (tracking?) start pixel

Stockman CSE803 Fall 2008 26

Outside material
 More examples of raster scanning
 Union-find algorithm and parent table
 Computing features from labeled object
region
 More on recursion and C++

Stockman CSE803 Fall 2008 27

Computing features of regions

Can postprocess results of CC alg.

Or, can compute as pixels are
aggregated

Stockman CSE803 Fall 2008 28

Area and centroid

Stockman CSE803 Fall 2008 29

Second moments

These are invariant to object location in the image.

Stockman CSE803 Fall 2008 30
 Contrast second moments

 For the letter ‘I’

 Versus the letter ‘O’ r
 Versus the underline ‘_’
c

Stockman CSE803 Fall 2008 31

Perimeter pixels and length

Stockman CSE803 Fall 2008 32

Circularity or elongation

Stockman CSE803 Fall 2008 33

Circularity as variance of “radius”

Stockman CSE803 Fall 2008 34

Radial mass transform

• for each radius r, accumulate the mass at distance r from

the centroid (rotation and translation invariant)
• can iterate over bounding box and for each pixel, compute a
rounded r and increment histogram of mass H[r]
Stockman CSE803 Fall 2008 35
Interest point detection
 Centroids of regions can be interesting
points for analysis and matching.
 What do we do if regions are difficult
to extract?
 We might transform an image
neighborhood into a feature vector, and
then classify as “interesting” vs “not”.

Stockman CSE803 Fall 2008 36

Slice of spine MRI and interesting
points selected by RMT & SVM

Stockman CSE803 Fall 2008 37

3D microvolumes from Argonne
high energy sensor: 1 micron
voxels

Ram CAT slice of a bee stinger (left) versus segmented slice

(right). Each voxel is about 2 microns per side.
Stockman CSE803 Fall 2008 38
 Scanning technique used
CCD camera material sample X-rays
scintillator

Pin head
X-rays partly absorbed by sample;
excite scintillator producing image in
the camera; rotate sample a few rotate
degrees and produce another image;
3D reconstruction using CT
Stockman CSE803 Fall 2008 39
 Different view of stinger

Rendered using ray tracing and pseudo coloring based on the

material density clusters that were used to separate object from
background. (Data scanned at Argonne National Labs)
Stockman CSE803 Fall 2008 40
Section of interesting points from
RMT&SVM

Stockman CSE803 Fall 2008 41

Segmentation of Scutigera

Stockman CSE803 Fall 2008 42

Scutergera: a tiny crustacean
• organism is smaller than

1 mm
• scanned by
• volume segmented and
meshed by Paul Albee
• roughly ten million triangles
to represent the surface
• anaglyph created for 3D
visualization (view with glasses)
Stockman CSE803 Fall 2008 43
Axis of least inertia
• gives object oriented
coordinate system
• passes through centroid
• axis of most inertia is
perpendicular to it
• concept extends to 3D
and nD

Stockman CSE803 Fall 2008 44

Derive the formula for best
axis
 use least squares to derive the tangent
angle of the axis of least inertia
 express tan 2 in terms of the 3
second moments
 interpret the formula for a circle of
pixels and a straight line of pixels

Stockman CSE803 Fall 2008 45