OTSU Thresholding Method for Flower Image Segmentation

ISSN (e): 2250 – 3005 || Volume, 06 || Issue, 05||May – 2016 ||
International Journal of Computational Engineering Research (IJCER)
www.ijceronline.com Open Access Journal Page 1
OTSU Thresholding Method for Flower Image Segmentation
Amruta B. Patil 1
, J.A.shaikh2
1
Electronics Engineearing, PVPIT ,Budhgaon, Sangli,M S, India.
2
Electronics Engineearing, PVPIT ,Budhgaon, Sangli,M S, India.
I. Introduction
Segmentation technique subdivides an image into different parts. It is a high level task which gives variety of
applications including object recognition, scene analysis or image/video indexing[1]. Image segmentation refers
to the process of partitioning a digital image into multiple segments i.e. set of pixels, pixels in a region are
similar according to some homogeneity criteria such as color, intensity or texture, so as to locate and identify
objects and boundaries in an image [2]. Thousands of different segmentation techniques are present in the
literature, but there is not a one method which can be considered better for different images, all methods are not
equally good for a particular type of image [3].Following are the different steps for the proposed method.
II. Proposed segmentation schema
Normally flower area contain a large green area covered by leaves and the remaining area is of flower which
also occupied by its color.The color is discriminative and canbe used as a homogeneity criterion to execute
foreground/ background segmentation [8]. In the first step we are converting RGB image into Lab color space.
Pre processing has done ,median filter is used for removal of noise present in the flower image.Applying OTSU
thersholding on each Lab component separeatly.Finally,by choosing the best result and applying the
postprocessing step for removing the all small regions from the background depending upon the threshold value
segmentation result will be obtained. Oxford data set is used for this experiment.some of which shown in Fig.2.
RGB
Image
Lab
Image
L
component
a
component
b
component
Pre
processing
Pre
processing
Pre
processing
OTSU
OTSU
OTSU
Evaluation
of best
result
Post
processing
Final
Segmentation
Result
Fig.1.Steps for proposed method
ABSTRACT
Segmentation is basic process in image processing. It always produces an effective result for next
process. In this paper, we proposed the flower image segmentation. Oxford flower collection is used for
segmentation.Different segmentation techniques are available. Different techniques and algorithm are
developed to describe the segmentation.We proposed a OTSU thresholding technique for flower image
segmentation in this paper. which gives good result as compared with the other methods and simple
also.Segmentation subdivide the image into different parts.firstly, segmentation techniques and then
otsu thresholding method described in this paper.CIE L*a*b color space is used in thresholding for
better results.Thresholding apply seperatly on each L, a and b component. accordingly the features can
be extracted like shape, color, texture etc. finally, results with the flower images are shown.
Keywords: Otsu thresholding, segmentation, CIE Lab color space, Region based segmentation, globle
thresholding method.

OTSU Thresholding Method For Flower…
Fig.2. Oxford flower collection
III. Segmentation techniques
Segmentation subdivides an image into its constituent region or object. Image segmentation methods are
categorized on the basis of two properties discontinuity and similarity [2]. Based on this property image
segmentation is categorized as Edged based segmentation and region based segmentation. The segmentation
methods that are based on a discontinuity property of pixels are considered as boundary or edges based
techniques. Edge based segmentation contains two methods gradient based and gray level histogram method,
while region based segmentation contain thresholding, region growing and region splitting and merging [4].
3.1. Thresholding
Thresholding is the simplest segmentation method. Thresholding process convert a multilevel image into a
binary image i.e., it select a proper threshold T, to divide image pixels into different regions and split objects
from background based on their level distribution. Thresholding creates binary images from Gary-level ones by
turning all pixels below some threshold to zero and all pixels about that threshold to one.Their are different
types of thresholding techniques.
i) Global thresholding, using an appropriate threshold T:
g(x,y) =
1 𝑖𝑓 𝑓 𝑥, 𝑦 > 𝑇
0, 𝑖𝑓 𝑓 𝑥, 𝑦 ≤ 𝑇

ii) Variable thresholding, if T can change over the image.
a) Local or regional thresholding, if T depends on a neighborhood of (x, y).
b) Adaptive thresholding, if T is a function of (x, y).
iii) Multiple thresholding:
g(x, y) =
𝑎, 𝑖𝑓 𝑓 𝑥, 𝑦 > 𝑇2
𝑏, 𝑖𝑓 𝑇1 < 𝑓 𝑥, 𝑦 ≤ 𝑇2
𝑐, 𝑖𝑓 𝑓 𝑥, 𝑦 ≤ 𝑇1
IV. Otsu Thresholding
It is important in picture processing to select an adequate threshold of gray level for extracting objects from their
background. Otsu is an automatic threshold selection region based segmentation method.Otsu method is a type
of global thresholding in which it depends only on gray value of the image. Otsu method was proposed by
Scholar Otsu in 1979. Which is widely used because it is simple and effective [5]. The Otsu method requires
computing a gray level histogram before running. However, because of the one-dimensional which only
consider the gray-level information, it does not give better segmentation result. So, for that two dimensional
Otsu algorithm was proposed which works on both gray-level threshold of each pixel as well as its Spatial
correlation information within the neighborhood. This algorithm can obtain satisfactory segmentation results
when it is applied to the noisy images [6]. Otsu’s method is expected in finding the optimal value for the global
threshold. It is based on the interclass variance maximization.
4.1 Formulation
Considering, the pixels of a given picture be represented in L gray levels [1, 2,…,L]. The number of pixels at
level i is denoted by ni and the total number of pixels by N = n1 + n2 + + nL.In order to simplify the discussion,
the gray-level histogram is normalized and regarded as a probability distribution [7] :
𝑝𝑖 = 𝑛𝑖 / 𝑁, 𝑝𝑖 > 0 , 𝑃𝑖 = 1𝐿
𝑖=1
We divide the pixels into two classes CO and C1 (background and objects, or vice versa) by a threshold at level
k; CO denotes pixels with levels [1, , k], and C1 denotes pixels with levels [k + 1,…. , L]. Then the probabilities
of class occurrence and the class mean levels, respectively, are given by
𝜔0 = 𝑃𝑟 (𝐶𝑜) = 𝑝𝑖 = 𝜔(𝑘)𝑘
𝑖=1
𝜔1 = 𝑃𝑟 (𝐶1) = 𝑝𝑖
𝐿
𝑖=𝐾+1
= 1 − 𝜔(𝑘)
and
𝜇0 = 𝑖𝑘
𝑖=1 Pr ( i | C0 ) =
𝜇 𝑘
𝜔(𝑘)
𝜇1 = 𝑖 Pr 𝑖 𝐶1)𝐿
𝑖=𝑘+1 = 𝜇 𝑇− 𝜇 𝑘 / 1 − 𝜔(𝑘)
where
𝜔 𝑘 = 𝑝𝑖
𝑘
𝑖=1
and
𝜇 𝑘 = 𝑖𝑝𝑖
𝑘
𝑖=1
which are the zeroth and the first-order increasing moments of the histogram up to kth level, and

𝜇 𝑇 = 𝜇 𝐿 = 𝑖𝑝𝑖
𝐿
𝑖=1
This is the total mean level of the original picture.We can verify for any value of k :
𝜔0 𝜇0 + 𝜔1 𝜇1 = 𝜇 𝑇 , 𝜔0 + 𝜔1 = 1
The class variance is given by,
𝜎0
2
= (𝑖 − 𝜇𝑘
𝑖=1 0
)2
Pr 𝑖 𝐶0) = (𝑖 − 𝜇0)𝑘
𝑖=1
2
𝑝𝑖/𝜔0
𝜎1
2
= (𝑖 − 𝜇𝐿
𝑖=𝑘+1 1
)2
Pr 𝑖 𝐶1) = (𝑖 − 𝜇1)𝐿
𝑖=𝑘+1
2
𝑝𝑖/𝜔1
These required second order cumulative moments. To measure the class separability at threshold level k
𝜆 = 𝜎 𝐵
2
/ 𝜎 𝑊
2
, 𝑘 = 𝜎 𝑇
2
/ 𝜎 𝑊,
2
, 𝜂 = 𝜎 𝐵
2
/ 𝜎 𝑇
2
where,
𝜎 𝑊
2
= 𝜔0 𝜎0
2
+ 𝜔1 𝜎1
2
𝜎 𝐵
2
= 𝜔0 ( 𝜇0 − 𝜇 𝑇)2
+ 𝜔1(𝜇1 − 𝜇 𝑇)2
= 𝜔0 𝜔1 (𝜇1 − 𝜇0)2
and
𝜎 𝑇
2
= (𝑖 − 𝜇 𝑇)𝐿
𝑖=1
2
𝑃𝑖
are the within class variance, the between-class variance, and the total variance of levels, respectively. well
threshold clasess would be separated in gray levels, and this threshold is the best threshold.
𝜎 𝑊
2
+ 𝜎 𝐵
2
= 𝜎 𝑇
2
It shows 𝜎 𝑇
2
is independent of k.but the fuction of 𝜎 𝑊
2
and 𝜎 𝐵
2
.It also shows that 𝜎 𝑊
2
is based on the second-order
statistics (class variances),while 𝜎 𝐵
2
is based on the first-order statistics (class means).
The optimal threshold k* that maximizes 𝜂 , or equivalently maximizes 𝜎 𝐵
2
is selected in the following
sequential search by
using the simple cumulative quantities.
𝜂 𝑘 = 𝜎 𝐵
2
𝑘 / 𝜎 𝑇
2
𝜎 𝐵
2
𝑘 =
[ 𝜇 𝑇 𝜔 𝑘 − 𝜇(𝑘)]
𝜔 𝑘 [1−𝜔 𝑘 ]
2
and the optimal threshold k* is,
𝜎 𝐵
2
𝑘 ∗ = 𝑚𝑎𝑥1≤𝑘<𝐿 𝜎 𝐵
2
𝑘
from the problem ,the range of k over which the maximum is sought can be restricted to
S* = 𝑘; 𝜔0 𝜔1 = 𝜔 𝑘 1 − 𝜔 𝑘 > 0, 𝑜𝑟 0 < 𝜔 𝑘 < 1
We consider it as a effective range of the gray-level histogram, always take into account the maximum threshold
value.

V. Experimental result
The evaluation was performed using a flower dataset provided by Oxford University which contains 17 spices
of flower having 840 images.some practical results and their threshold values are shown in above figure
original Image L component a component b component
Threshold value for L component = 167
a component = 109
b component = 189.
So b component is having good result rather than L and b in this case.
a) original image b) L component c) a component d) b coponent
Fig.3. RGB to Lab conversion output
VI. Conclusion
In this scenario, we proposed a fast flower segmentation depending upon OTSU thersholding and Lab color
space which gives good result. The results are depend on threshold value of each component of image i.e Lab
component. Applying OTSU on the three components separately gives good result as compared to other
methods. In some cases lost of information which will be overcome by modifying the OTSU algorithm. The pre
and post segmentation benifitial to remove the noise. Due to fine segmentation it would be easy to apply feature
extraction schema like color,texture,and shape on segmented image. This would be the further work for this
process.

References
[1] A Learning Approach for Adaptive Image Segmentation, Vincent Martin and Monique Thonnat INRIA Sophia Antipolis,
ORION group,France
[2] Rafael C. Gonzalez, Richard E. Woods, “Digital Image Processing”, 2nd ed., Beijing: Publishing House of Electronics
Industry, 2007.
[3] K. K. Singh, A. Singh,“A Study of Image Segmentation Algorithms for Different Types of Images”, International
Journal of Computer Science Issues, Vol. 7, Issue 5, 2010.
[4] A. B. Patil, J.A.Shaikh,“Segmentation And Feature extraction of flowers for Image Retrieval : A survey” ,IJARECE, Volume 5,
Issue 1, January 2016
[5] Zhong Qu and Li Hang”Research on Image Segmentation Based on the Improved Otsu Algorithm.”, 2010
[6] LIU Jian-Zhuang, Li Wen-Qing, “The Automatic threshold of gray level pictures via Two-dimensional Otsu Method”,
Acta Automatic Sinica, 1993
[7] A Tlreshold Selection Method from Gray-Level Histograms, IEEE Transactions on systems, vol., no 1, Jan. 1979.
[8] Asma Najjar et.al., “ Flower image segmentation based on color analysis and a supervised evaluation”, 2012 IEEE.
Ms. A. B. PATIL has received her Bachelor’s degree &
Diploma in Electronics And Telecommunication Engg. from
Shivaji University. Currently pursuing Master’s in
Electronics Engg. from Shivaji University, Kolhapur. Also
having 2 year Industrial Experience.
Mr .J. A. Shaikh (M.E.Electronics) Ph.D.(pursuing)
working as an Associate Professor & H.O.D at PVPIT,
Budhgaon. Having 24 Years of Teaching Experience.
His area of specialization is Power Electronics and
Image Processing.

OTSU Thresholding Method for Flower Image Segmentation

More Related Content

What's hot (17)

Similar to OTSU Thresholding Method for Flower Image Segmentation (20)

Recently uploaded (20)

OTSU Thresholding Method for Flower Image Segmentation