You are on page 1of 16

Newer Image Segmentation Methods for

Biomedical Object Detection


Tarundeep Singh Dhot
Dept of ECE
Concordia University
Montreal, QC H3G 1M8

ABSTRACT image processing since it is used to isolate


The basic aim of the study is intended towards physiological and biological structures of
the field of image segmentation of cells with the interest. General approaches to segmentation
aim that it can be further extended to detect cell can be grouped into three classes: pixel-based
inclusions. The study can be divided into three methods, regional methods and edge-based
phases. The first phase deals with the methods.
segmentation of individual cell from a cell Traditionally, pixel-based methods are the
image. Three segmentation approaches have easiest to understand and to implement, but are
been used for this purpose. This phase is also the least powerful and, since they operate
followed by image mapping of the cell into on one element at a time, they are particularly
various image intensity plots. It is then intended susceptible to noise. Continuity-based and edge-
to use a Genetic Algorithm to locate the global based methods approach the segmentation
optima of the plot. problem from opposing sides: edge-based
methods search for differences while continuity-
based methods search for similarities.
Categories and Subject Descriptors
I.4 [Image Processing and Computer In this study, three segmentation approaches
Vision]: Segmentation – pixel classification have been used: pixel-based methods,
continuity-based multi-thresholding and the
General Terms third one is based on a recent research
Algorithms, Experimentation publication “Automatic segmentation of Cells
from Microscopic Imagery using Ellipse
Keywords Detection”. It is also important to mention the
Image Segmentation, Genetic Algorithms use of morphological operators in the last
approach. Morphological operations have to do
with processing shapes. In this sense they are
continuity-based techniques, but in some
1. INTRODUCTION applications they also operate on edges, making
them useful in edge-based approaches as well.
The field of Biomedical Image Processing is an
exciting and growing cross-disciplinary field. After loading the image, pre-processing of the
An exponential increase in the power of digital image was carried out in terms of reduction of
processors, a steady growth in the field of noise using various filters.
Digital Image Processing principles has given a
much needed impetus for emerging as a mature Once the image is pre-processed and
subject in its own rights. One of the most successfully segmented, mapping of the image
upcoming and active areas of interest is the of the individual cell is done into various image
analysis of cell images for detection of cell intensity plots. These intensity plots of the
inclusions. Cell image segmentation is a image represent the segmented image into
necessary first step of many automated various 3-Dimensional surface plots of their
Biomedical Image Processing procedures. matrix data. These intensity plots show exactly
the location of the cellular inclusion on the basis
Image segmentation is the identification and of presence of a region (inclusion) having
isolation of an image into regions that one similar pixel values. Thus, the concentration of
hopes, correspond to structural units. It is an the pixel values is represented in the intensity
essentially important operation in biomedical plot over the entire cellular region. Once the
segmented image is represented in as a 3-D 2.1 Approach 1: Pixel-Based
intensity plot, the last intended step is to use a Segmentation Using Minimal
genetic algorithm to locate the global optima.
This phase is currently in the process of Variance Iterative Technique For
formulation. Thresholding

The project is developed on the MATLAB 7.0 The most straightforward and common of the
platform. pixel-based methods is thresholding in which all
pixels having intensity values above, or below,
some level are classified as part of the segment.
Thresholding is an integral part of converting an
2. METHODOLOGY intensity image to
a binary image. Thresholding is usually quite
As stated above, the study has three phases: fast and can be done in real time allowing for
1. Segmentation of individual cells from cell interactive setting of the threshold. The basic
image; concept of thresholding can be extended to
2. Mapping image of individual cell to an include both upper and lower boundaries
intensity image; and (slicing).
3. Use of GA to locate the global maxima
A major concern in pixel-based methods is
setting the thresholding or slicing level(s)
Let us start of with cell segmentation. appropriately. Usually these levels are set by the
In segmentation, the inputs are images but the program, although in some situations they can
outputs are attributes extracted from those be set interactively by the user. Finding an
images i.e. segmentation subdivides an image appropriate threshold level can be aided by a
into its constituent regions or objects. In the plot of pixel intensity distribution over the
study, three approaches to cell segmentation are whole image, regardless of whether you adjust
used. We shall see them individually in detail. the pixel level interactively or automatically.
Such a plot is termed as the Intensity
The first step in cell segmentation is loading the Histogram. Intensity histograms can be very
image and converting it to gray scale from the helpful in selecting threshold levels, not only
usual RGB color format. The image is then for the original image, but for images produced
further pre-processed. This step is common to by various segmentation algorithms. Initially
all the methods used. This step can be referred histograms can be useful in evaluating the
to as the Loading Subroutine. efficacy of different processing schemes: as the
separation between structures improves,
Loading Subroutine: histogram peaks should become more
distinctive. If the intensity histogram is, or can
a) Load the image be assumed as bimodal (or multi-modal), a
b) Perform grayscaling. common strategy is to search for low points, or
c) Improve contrast for the grayscaled minima, in the histogram.
image.
d) Improve precision (of the image from Keeping this in mind, there are two approaches
step c). based on which Approach 1 is designed and
implemented.
It may be remembered that the images are all in
the bitmap format (.bmp). One inclusion from The first approach is to improve the
the slide was pasted on a black background. determination of histogram minima is based on
This also reduces the effect of noise. All the observation that many boundary points carry
segmentation techniques are implemented on values intermediate to either side of the
this image to show variance in effectiveness of boundary. These intermediate values will be
the three methods. associated with the region between the actual
boundary values and may mask the optimal
threshold value. However, these intermediate
points have the highest gradient, and it should
be possible to identify them using a gradient-
sensitive filter, such as the Sobel or Canny on the final segmented image in order to plot
filter. After these boundary points are identified, the intensity graphs.
they can be eliminated from the image, and a Figure 1.1 displays broadly the various image
new histogram is computed with a distribution transformations of Pixel-based segmentation
that is possibly more definitive. Thus, this leads approach used for Approach 1. The original
to slightly better segmentation of the cell. image as shown in Fig 1.1 a) is first converted
to an increased contrast, high precision gray
The second threshold strategy is one that does scale image as shown in Fig 1.1 b). Firstly, edge
not use the histogram is based on the concept of pixels are removed and then the image is
minimizing the variance between presumed thresholded to give the final segmented image
foreground and the background elements. shown in Fig 1.1 c). Intensity plot are
Although the approach assumes two different determined on Fig 1.1 c) which gives rise to the
gray levels, it works well even when the intensity plot as shown in Fig 1.1 d) which is a
distribution is not bimodal. The approach uses meshed image intensity plot.
an iterative process to find a threshold that
minimizes the variance between the intensity 2.2 Approach 2: Image Segmentation
values on either side of the threshold level Using Multi-Thresholding Technique
(Outso’s method). This can be referred to as
Minimal Variance Iterative Technique. (Using AND - OR Operators)

The above two approaches are thus incorporated The results of several different segmentation
and implemented for Approach 1 as follows: approaches can be combined either by adding
the images together or more commonly, by first
1. Load the image of the cell and perform the thresholding images into separate binary images
necessary pre-processing like gray scaling and then combining them using logical
on the image (Loading Subroutine) operations. Either the AND or OR operator
2. Display images after pre-processing along would be used depending on the characteristics
with the intensity histogram of each segmentation procedure. If each
3. Remove the edge pixels from the image and procedure identified all of the segments, but
display the histogram of this modified also included non-desired areas, the AND
image. operator could be used to reduce the artifacts.
4. Determine thresholds using minimal Alternatively, if each procedure identified some
variance iterative technique described portion of the segment(s), then the OR operator
above could be used to combine the various portions.
5. Apply this approach to threshold both
images. This approach is used in Approach 2 where first
6. Display the resultant thresholded images. two, then three, thresholded images are
7. Perform class conversion of the final combined to improve segment identification.
segmented image The structure of interest is a cell which is shown
8. Plot the image intensity plots on a gray background. Threshold levels above
and below the gray background are combined
To remove the edge boundaries, first (after one is inverted) to provide improved
identify these boundaries using an edge isolation. Including a third binary image
detection scheme. While any of the edge obtained by thresholding a texture image further
detection filters can be used, for our approach, improves the identification.
Canny filter is used as it is more robust to noise.
The implementation of this filter will produce a Implementation of Approach 2:
binary image of the boundaries. This boundary
image is converted to a boundary mask by 1. Load the image of the cell and perform the
inverting the image. After inversion, the edge necessary pre-processing like gray scaling
pixels will be zero while all other pixels will be on the image (Loading Subroutine)
one. Multiplying the original image by the 2. Display images after pre-processing
boundary mask will produce an image in which 3. Perform linear filtering of the image using
the boundary points are removed (i.e. set to zero Low Pass Filter
or black). Perform class conversion (to double - 4. Threshold the image its complement
double-precision floating-point number array) 5. Threshold texture image
6. Combine the thresholded image and its
complement
7. Combine the above image with the
threshold texture image
8. Display thresholded and combined images
9. Perform class conversion of the final
segmented image
Fig 1.1 a) Original Image 10. Plot the image intensity plots

It is possible to isolate some portions of the cell


by thresholding above and below the
background level. After inversion of the
thresholded image, the images are combined
using a logical OR. Since the cell also shows
some textural features, a texture image is
constructed by taking the regional standard
deviation. After thresholding, this texture-based
image is also combined with the other two
Fig 1.1 b) High Precision Gray images.
Scaled Image
The figure on the left displays the main image
transformation of the cell using the Multi-
Thresholding technique used in Approach 2.
Again, here too, first the original image shown
in Fig 1.2 a) is pre-processed into an increased
contrast, high precision gray scale image shown
in Fig 1.2 b). The image shown in Fig 1.2 c) is
obtained after use of logical OR operator on the
thresholded and inverse thresholded image. Fig
1.2 d) shows the logical OR operation of image
Fig 1.1 c) Segmented Image (Thresholded obtained in Figure 1.2 c) with the textural
Edge Removed) image.
Fig 1.2 e) and 1.2 f) show the intensity plots for
Fig 1.2 c) and d).

2.3 Approach 3: Automatic


Segmentation of Cells from
Microscopic Imagery using Ellipse
Detection:

In this approach, a two-phase algorithm that


combines iterative thresholding with ellipse
detection is used for potent cell segmentation
mechanism. In this approach, a combination of
Morphological Operators with adaptive
thresholding is used in order to simultaneously
eliminate noise and locate inclusion regions,
hence, computing the best possible local
Fig 1.1 d) Image Intensity Plot (Mesh)
threshold levels for the inclusion extraction.
Figure 1.1: Image transformations using
Pixel –Based Segmentation The main reason why this approach of also
included in this study was due to the use of
morphological operators. As stated earlier,
Fig 1.2 a) Original Image

Fig 1.2 e) Intensity Plot

Fig 1.2 b) High Precision Gray


Scaled Image

Fig 1.2 f) Intensity Plot

Figure 1.2: Image transformations using


Multi –Thresholding

Fig 1.2 c) Combined Thresholded Image


morphological operations have to do with
processing shapes. The two most common
morphological operations are dilation and
erosion. In dilation the rich get richer and in
erosion the poor get poorer. Specifically, in
dilation the centre or active pixel is set to the
maximum of its
neighbors and in erosion it is set to the
minimum of its neighbors. Since these
operations are often performed on binary
images, dilation tends to expand edges, borders
or regions, while erosion tends to decrease or
Fig 1.2 d) Combined Thresholded and even eliminate small regions. The two processes
Textural Image can be done in tandem, over the same area.
Since erosion and dilation are nonlinear
operations, they are not invertible
transformations i.e. one followed by the other
will not generally result in the original image. If
erosion is followed by dilation, the operation is
termed as opening. If image is binary, this
combined operation will tend to remove small image (Fig 2 h). The intensity plot (mesh) of
objects without changing the shape and size of this image is shown in Fig 2 (i).
large objects. Basically, the initial erosion tends
to reduce all objects, but some of the smaller Fig 2.1 deals with the various intensity plots
objects will disappear altogether. The depending on the function against which the
subsequent dilation will restore those objects image is plotted. Three such functions are
that were not eliminated by erosion. If the order shown: mesh, ribbon and surf.
is order is reversed and dilation is performed
first followed by erosion, the combined process Also displayed are two pairs of histograms (Fig
is called closing. Closing connects objects that 2.2 and 2.3), the first one comparing a simple
are close to each other, tends to fill up small gray scale image to a high precision image, and
holes, and smooths an object outline by filling the second one showing improved separability
small gaps. As with the more fundamental when boundaries are eliminated.
operations of dilation and erosion, the size of
objects removed by opening and closing Similarly, the output images of Approach 2 i.e.
depends on the size and shape of the Multi-Thresholding are displayed in Fig 3. One
neighborhood that is selected. thing to note here is that Fig 3 (e) shows the
Thresholded image. This image is then inversed
Implementation of Approach 3: (complemented) which is shown in Fig 3 (f) and
these both images are then logically ORed to
1. Load the image of the cell and perform the produce the Combined Threshold Image of Fig
necessary pre-processing like gray scaling 3 (g). This image is again ORed with a textural
on the image (Loading Subroutine) image shown in Fig 3 (h) to give rise to
2. Display images after pre-processing Combined Threshold Textural Image shown in
3. Perform linear filtering of the image using Fig 3 (i). Fig 3 (j) shows the intensity plot of
Low Pass Filter only image Fig 3.1 (g) while Fig 3 (k) is the
4. Application of Open Morphological intensity plot of Fig 3 (j).
Operation:
Erode  Dilate. Fig 3.1 represents the various intensity plots.
5. Global Thresholding.
6. Multi-Region Adaptive Thresholding. Fig 4 represents the output images of Approach
7. Second application of Open Morphological 3: Auto-segmentation of cells. Fig 4 (e) and (f)
Operation represent the first operation of the
8. Display thresholded images morphological operators: Erode followed by
9. Perform class conversion of the final Dilation i.e. opening. Fig 4 (g) represents image
segmented image of cell after first thresholding. Similarly, Fig 4
10. Plot the image intensity plots (h) and (i) represent the “second opening”
operation while Fig 4 (j) represents the second
3. RESULTS thresholding on the cell.

The results for the study are described in the Fig 4.1 show the intensity plots after each
form on the images obtained from the opening operation. Fig 4.2 shows the various
experiment. Image output obtained from all the intensity plots for the final thresholded image in
three approaches are displayed in a sequential Fig 4 (j).
flow chart format for easier understanding. The
study was carried out on an Intel Pentium 4, 4. DISCUSSION
3.20GHz, 512 MB DDR RAM. Experiments
were performed on MATLAB 7.0. In this project, an effort was made to study three
different segmentation approaches even though
As shown in Fig 2 for Pixel-based edge no broad comparison study was made. As it can
removed thresholding, after the initial pre- be seen, the output segmented images of all the
processing of the images, using a Canny filter, three methods are completely different. This
the edges are removed (Fig 2 g). This image is shows how segmentation results from different
then thresholded to give the final segmented approaches can yield to completely different
that pixel based methods are particularly
Fig 1.3 e) Second Opening
Fig 1.3a) Original Image

Fig 1.3 f) Final Thresholding


Fig 1.3b) High Precision Gray
Scaled image

Fig 1.3c) First Opening Fig 1.3 g) Intensity Plot

Fig 1.3: Image transformation using


Automatic Segmentation Approach

susceptible to noise, since they operate on one


element at a time. Filtering can reduce the effect
of noise, thus is used in all the three approaches.

Also with Multi-Thresholding, we can logically


Fig 1.3 d) First Thresholding AND or OR images according to our
requirement which can be really helpful. Use of
textured image also increases the options for
use. Lastly, morphological operators allow for
manipulation of maxima and minima of an
image. Also with intensity plots, visualization
of cell inclusions improves as it shows
magnitude of relative intensity to its location.
5. FIGURES

5.1 Pixel Based Methods


5.1.1 Phases in Image Transition

a) Original Image b) Gray Scaled c) Increased Contrast

h) Segmented Image:
Thresholded Edge
d) High Precision
Removed
i) Intensity Plot (Mesh)

e) Gaussian Low
g) Edge Removed f) 3-D Convolution
Filter (Canny) Pass Filter

Fig 2: Pixel-Based Segmentation Using Minimal Variance Iterative Technique For


Thresholding
5.1.2 Intensity Plots

a) 3-D Mesh Plot

b) 3-D Ribbon Graph of Matrix

c) 3-D Surface Plots

Fig 2.1 Various Plots of Image Intensity for Pixel Based Edge-Removed Thresholding
5.1.3 Histograms

Fig 2.2: Histograms of Gray Scaled Image and High Precision Image. There is
significant improvement in the histogram of High precision Image
Fig 2.3: Images of cell with (upper) and without (lower) intermediate boundaries
removed. The associated histograms (right side) show improved separability when
boundaries are eliminated
5.2 Multi-Thresholding

5.2.1 Phases In Image Transition

a) Original Image b) Gray Scaled c) Increased Contrast d) High Precision

e) Thresholded f) Complement
Image of Threshold

j) Intensity Plot: Combined Threshold


g) Combined Threshold h) Textural Image
Image
Image

j) Intensity Plot: Combined Threshold


i) Combined Threshold
Textural Image Textural Image
Fig 3: Image Segmentation Using Multi-Thresholding Technique (Using AND
- OR Operators)
5.2.2 Intensity Plots

(a) Intensity Plots of Combined Thresholded and Threshold Textural Image

(b) Intensity Represented by Various Functions

Fig 3.1: Intensity Plots: Multi-thresholding Technique


5.3 Auto-Segmentation Of Cells
5.3.1 Phases In Image Transition

a) Original Image b) Gray Scaled c) Increased Contrast d) High Precision

e) Erode - I

j) Intensity Plot (Mesh and Ribbon)


f) Dilate - I

k) Threshold - II i) Dilate - II h) Erode - II g) Threshold - I


Fig 4: Automatic Segmentation of Cells
5.3.2 Intensity Plots Of Morphological Operations

a) Mesh Plot After Erosion - I b) Mesh Plot After Dilation - I

c) Mesh Plot After Erosion - II d) Mesh Plot After Dilation - II


Fig 4.1 Intensity Plots After Opening Operations

The above figure shows various intensity plots after the use of “Opening Operation” before and after
thresholding of the image. The marked change in the plots can be seen. The plots before thresholding suffer
from presence of noise in the image. After thresholding, presence of noise seen is minimal.
5.3.3 Intensity Plots

(c) 3-D Contour Plot of Image


(a) 3-D Mesh Plot of Image (b) 3-D Surface Plot of
Image

Fig 4.2: Intensity Plots for Automatic Cell Segmentation Method

You might also like