OBJECT DETECTION, EXTRACTION AND CLASSIFICATION USING IMAGE PROCESSING TECHNIQUE

Journal for Research | Volume 03 | Issue 08 | October 2017
ISSN: 2395-7549
All rights reserved by www.journal4research.org 9
Object Detection, Extraction and Classification
using Image Processing Technique
Ms. Payal Bose Prof. Samir Kumar Bandyopadhyay
M.Tech Student Professor
Department of Computer Science & Engineering Department of Computer Science & Engineering
University of Calcutta University of Calcutta
Abstract
Domestic refrigerators are widely used household appliances and a large extent of energy is consumed by this system. A phase
change material is a substances that can store or release significant amount of heat energy by changing the phase liquid to vapour
or vice versa. So, reduction of temperature fluctuation and improvement of system performances is that main reason of using
PCM enhances the heat transfer rate thus improves the COP of refrigeration as well as the quality frozen food. The release and
storage rate of a refrigerator is depends upon the characteristics of refrigerators and its properties using phase change material for
a certain thermal load it is found that COP of conventional refrigerator is increased . The phase change material used in chamber
built manually and which surrounds the evaporator chamber of a conventional refrigerator the whole heat transfer for load given
to refrigerator cabin (to evaporator) evaporator to phase change material by conduction. This system hence improves the
performances of household refrigerator by increasing its compressor cut-off time and thereby minimizing electrical energy usage.
The main objective is to improve the performance, cooling time period, storage capacity and to maintain the constant cooling
effect for more time during power cut off hours using phase change material.
Keywords: Edge Detection, segmentation, classification, Object extraction and Object Detection
_______________________________________________________________________________________________________
I. INTRODUCTION
Object extraction is a challenging area in the image processing. In object extraction the technique of extracting objects from the
pre-processed image is done in such a way that similarity class within image is made into a number of clusters for isolating
segmented images from the original image. The nontrivial contents, usually in the form of interesting objects, are sufficient to
represent the semantic meanings in most cases and consequently play an important role in many image applications such as
content-based retrieval. Therefore, many methods have been proposed to automatically extract interesting objects. Image Object
extraction techniques are very useful for disease detection, object localization and object tracking. The method used for still
image object extraction can also use for 3D image and video frames for the same purpose. Object Extraction is a closely related
issue with the segmentation process. Image Segmentation is a process of dividing an image into sub partition based on some
characteristics like color, intensity etc.
The main goal of object extraction is to change the representation of an image into something more meaningful. To extract an
object from the image first we have to segment the entire image. User select the region as background and foreground by using
the markers and then the algorithm will segment the image and the foreground region will be extracted from the image. Image
segmentation is a fundamental step in many areas of computer vision including object recognition, video surveillance, face
recognition, fingerprint recognition etc. It provides additional information about the contents of an image by identifying edges
and regions of similar color and texture. Although a first step in high level computer vision tasks, there are many challenges to
ideal image segmentation. Segmentation subdivides an object into its constituent regions or objects. The level of detail to which
the subdivision is carried on depends on the problem being solved. That is the segmentation should stop when regions or objects
of interest have been detected. For example, if an image consists of a tree, the segmentation algorithm may either stop after
detecting the entire tree or further divide the tree into trunk and leaves.
Interactive image segmentation is a way to extract foreground objects in complex scenes using simple user interaction. The
key to success in interactive image segmentation is to preserve characteristics of the user’s interactive information and maintain
global data effectively. 
II. REVIEW WORKS
Images contain different types of objects and structures which may convey information. counting involves estimating the number
of objects in an image, detecting involves presence the number of objects in an image. Counting arises in many real time
application such as counting grains in agriculture industry, counting cells in microscopic images, counting of number diamonds
in industry etc. Existing methods for counting involves a large amount of hardware which also adds to the cost or manual
counting which is time consuming and may give erroneous results [1-4]. In this paper image segmentation is stated as a vast
topic of research and choice of large number of researchers by the author. The reason for the popularity of image segmentation is

Object Detection, Extraction and Classification using Image Processing Technique
(J4R/ Volume 03 / Issue 08 / 003)
because of its importance in the area of image processing and computer vision. The prime task of the researchers working in the
field is to develop a method for efficient and better image segmentation. The segmentation done using approaches of clustering
are considered good for image segmentation. The advantage of using approaches of clustering in image segmentation is that this
is a wide area and can be employed in other areas of engineering too. In this paper the author has developed a new technique for
image segmentation keeping clustering as a base. K-mean algorithm is employed and distance parameter is considered for
deciding the performance. The distance measure „cosine‟ is employed in this paper. Sobel filter is then used for filtering and the
results are obtained using Marker Watershed algorithm. The performance parameters that are taken into consideration by the
author in this paper are Mean Square Error and PSNR [5-9]. In this paper the process of image segmentation is defined as the
technique via which we segment a given photograph into several parts in order that we can further analyzed every of these
components present in the photo. The author states that it is possible to extract some records via analyzing them and this statistics
is useful for excessive-stage gadget vision software. There are numerous techniques of photograph segmentation to be had in
literature. In this paper, analysis is done to examine the discontinuity-primarily based approach for photo segmentation. The
discontinuity-based totally segmentation may be categorised into 3 techniques: factor detection, line detection, and aspect
detection. The result of these numerous strategies is analyzed in MATLAB the use of IPT. The author additionally enforce the
unique part operators inclusive of Prewitt, Roberts, LoG, Canny and the consequences of these operators can be shown on
diverse pics[10-13]. In this paper the author offers a new method to picture segmentation the usage of Pillar okay-approach set of
rules and the algorithm defined using that set of rules is known as Pillar k-mean s algorithm. This segmentation method includes
a new mechanism for grouping the factors of high resolution pictures so that you can improve accuracy and decrease the
computation time. The system uses k-way for image segmentation optimized by means of the set of rules after Pillar. The Pillar
algorithm considers the location of pillars must be located as a long way from every other to face up to the pressure distribution
of a roof, as equal as the range of centroids between the information distribution. This set of rules is able to optimize the k-mean
clustering for photo segmentation in the aspects of accuracy and computation time. This set of rules distributes all initial
centroids in line with the most cumulative distance metric. In this paper a new technique for image segmentation is developed
that compares the results of K-mean algorithm with Gaussian aggregate model [4]. Experimental consequences make clear the
effectiveness of our approach to improve the segmentation satisfactory and accuracy factors of computing time . In this paper a
new histogram thresholding fuzzy C-method hybrid (HTFCM) approach is presented that would find distinct software in sample
popularity in addition to in laptop imaginative and prescient, particularly in shade photo segmentation. The histogram
thresholding approach that is proposed in the paper is employed to acquire all feasible uniform regions within the coloration
photograph. Then, the bushy C-manner algorithm is applied in the uniform regions while cluster formation and that will enhance
the compactness of formed clusters. Experimental outcomes have confirmed that the low complexity of the proposed HTFCM
technique should acquire better result[5].
III. PROPOSED METHOD
In this paper image processing using MATLAB commands are used to implement edge detection, segmentation, object detection
and object extraction of an image. For this problem, MATLAB graphical interface is used. Our aim is to accept an image from
user, then show its pixel value, convert it into grayscale and binary (black and white) image, also showing the pixel value of
grayscale image. Then enhancing the image. After that detect the edge of that image and based on it segment the image. After
segmentation detect the objects in the image and finally extract the objects from the image. These applications are also applicable
for medical images.
The stages in the methodology are shown below:
The steps involve in this methodology are describe below:
Accept Image (From User):
In this step, users are free to accept any type of image for performing the above next steps. It support file format JPG (.jpg) only.
The input image must be in RGB format. The input RGB image consists of the matrices of same size, representing the red green

(J4R/ Volume 03 / Issue 08 / 003)
and blue color of the image. This matrices are also says that how much color of each channel (red, green, blue) are used in the
image.
Converting into Greyscale:
Grayscale images consists of color grey, which consists different shades between black and white. After accepting input image
then converting it into greyscale represented a luminance image. It also creates the three matrices of red, green and blue channels
using linear intensity encoding by gamma expansion.
Converting into Binary:
Binary Image means an image only contains black and white color, it only assign two-pixel intensity value 0 and 1. It also called
monochromatic color. When an input image is converting into binary images, then it creates an image, which only consists of
black and white color. The value of three matrices corresponding red, green, and blue color is 0 (for black) and 1 (for white).
Image Enhancing:
There are two kinds of techniques used here for enhancing the image 1) Contrast stretching & 2) Histogram equalization. For
applying this method must convert the input image into grey image. The above techniques are only works on grey images.
In contrast stretching technique the intensity of an image is span to a desired range of values.
In histogram equalization technique the intensity of the image is adjust in such a way that it enhance the image contrast.
Edge Detection:
In edge detection technique, it finds the edges of the image based on the local maxima. First, convert the original RGB image
into greyscale image. Then filtering the all three color channel and resize the image. Then extracting the histogram values of the
filtering and try to find the local maxima based of the light reflection of the greyscale image pixel by pixel. After that finding the
local maxima, sort them and eliminate maxima value of certain range. Then using method multithresh(A,N) which
returns thresh a 1-by-N vector containing N threshold values using Otsu's method and use an input argument to imquantize to
convert image into an image with N + 1 discrete levels. After that, we get the edges of the whole image without any noise.
Segmentation:
Segmentation is an operation that partitioned an image into its component or it separate the objects. For segmentation edge
detection and thresholding is very important. After accepting, the input image from user must convert it into the greyscale image
and then for thresholding first consider a certain value in between 0 to 1 of each color channel (red, green and blue). Then
convert each channel into binary image and take the sum of three channel. Then complement the sum channel. Complement
means it convert the white space into black and vice versa, if there have any small holes then fill it with Ifill command.
Object Detection:
Object detection means determine how many objects are present in an image. So for this after segmented the input image find the
morphological structure of the input image, and then open the morphological image in greyscale mode. After that, find the area
and eccentricity of the all segmented areas and label them as well as bounded the all-region based on the connectivity with a
colourful bound box. After bounding all the objects count how many objects are in the input image.
Object Extraction:
Object Extraction means after segmentation and object detection just extracted all the objects from the images and show them all
individually. For this, after bounded all the regions label them. After labelling all the regions, find how many regions are present
in the images. After that using a loop for collecting or retrieving the objects from images. Here all the object images are in
greyscale mode.
For Medical Image:
Medical images are special images, they are RGB image but it’s not color image. Therefore, for thresholding it has not needed to
thresholding each color channel. Here consider a certain thresholding value, check the grey value of the image is greater than the
threshold value or not. Then take complement it, segment it, give boundary of the objects, and extract them.
IV. RESULTS
The results are shown in the following figures.

(J4R/ Volume 03 / Issue 08 / 003)
Accept Image (From User):
Fig. 1:
Here user chooses their input image. (Fig-1)
Converting into Greyscale:
Converting the input image (Fig-1) into greyscale (Fig-2a) and shown the histogram of greyscale image (Fig-2b).
Fig. 2a: Fig. 2b:
Converting into Binary:
Converting the input image (Fig-1) into greyscale (Fig-3)
Fig. 1: Fig. 3:

(J4R/ Volume 03 / Issue 08 / 003)
Image Enhancing:
Image Enhancing did by two ways 1) Contrast Stretching and 2) Histogram Equalization.
Contrast stretching:
The result of this technique is shown below: (Fig – 4a & 4b)
Fig. 4a: Fig. 4b:
Histogram Equalization:
The result of this technique is shown below: (Fig – 4c & 4d)
Fig. 4c: Fig. 4d:
Edge Detection:
In this technique our aim to find the boundary of all the objects in the input image (Fig – 1). The result figure (Fig – 5) shown
below.
Fig. 1: Fig. 5:

(J4R/ Volume 03 / Issue 08 / 003)
Segmentation:
In this technique our aim is to separate the objects of input image (Fig-1). The segmented image shown in (Fig- 6).
Fig. 1: Fig. 6:
Object Detection:
In this technique, our aim is to identify all the objects in the input image (Fig – 1) and bounded the objects. The resultant image
shown in (Fig – 7).
Fig. 1: Fig. 7:
Object Extraction:
In this technique, our aim to separate all the images from the input image (Fig -1) and display them separately. The resultant
images are shown below (Fig -8). Same kind of images is not shown below.
Fig. 1: Fig. 8:

(J4R/ Volume 03 / Issue 08 / 003)
For Medical Image:
The result of medical image after all operations is given below:
Fig. 9: Original Image Fig. 10: Binary image
V. ALGORITHM
The algorithms of the above methodology given below:
Algorithm for accepting image, various conversion, color channel analysis and image enhancing method
1) Accepting an input RGB image from user (File format is .jpg);
2) Display the input image;
3) Convert the image into greyscale;
4) Display the greyscale image;
5) Display the histogram image of the greyscale image;
6) Convert the input image into binary image(black and white)
7) Display the binary image
8) imR = read input image file
make imR(:,:,2,3) = 0 ; // it only show the red channel of RGB image, green and
blue channel is zero
display red channel image;

(J4R/ Volume 03 / Issue 08 / 003)
9) imG = read the input image file;
make imG(:,:,1:2:3) = 0; //only show the green channel, red and blue channel = 0
display green channel image;
10) imB = read input image file
make imB(:,:,1:2) = 0; //only show the blue channel, red and green channel = 0
display blue channel image;
//Contrast Stretching
11) Adjust the contrast of the greyscale image of input image
12) Display it;
13) Display the histogram image of adjusting image;
//Histogram Equalization
14) Adjust the histogram of the greyscale image of input image
15) Display it;
Algorithm for Edge Detection
1) Display it;
//Histogram Equalization
3) Adjust the histogram of the greyscale image of input image
4) Display it;
Algorithm for Edge Detection
1) Accept input from user
2) Convert it into greyscale image
3) K = filter the image by creating a filter array
4) Length_k, breath_k = size(k)
5) for index 1 to 255
Counts(index+1,1) = 0
end
// extracting histogram image
6) [Counts, X] = find the histogram image of image k
7) Calculate, total = length_k × breath_k
8) for index 1 to 255
calculate, p(index,1) = counts(index+1,1)/total
end
9) a_index = 0 //initialize
10) for index 1 to 254 // skip 1st
and last element
11) check , if counts(index,1)<counts(index+1,1) and counts(index+2,2)<counts(index+1,1)
then, a_index = a_index+1;
a(a_index,1) = index;
a(a_index,2) = counts(index+1,1);
end
end
12) a_size = size(a);
// find the maxima of maxima freq
13) fmax = 0; //initialize
14) for index 1to 255
15) Check, if a(index,2) >fmax
then, fmax = a(index,2);
end
end
// finding maxima that is within 0.01 of frequency maxima and store then into array b
16) b_index = 0 ; //initialize
17) for index 1 to a_index

(J4R/ Volume 03 / Issue 08 / 003)
check, if a(index,2) >= 0.01 × fmax
then, b_index = b_index+1;
end
end
// eliminating maxima within intensity range of 20 and storing in array c
18) c_index = 0; //initialization
19) c_int = b(1,1);
20) c_freq = b(1,2);
21) Index = 1
22) While index<b_index
While b(index+1,1) – c_int <=20
Check,if b(index+1,2) > c_freq
then, c_int = b(index+1,1);
c_freq = b(index+1,2);
end
again check, if(index+1)<b_index
then, index = index+1;
else
break;
end
end
Calculate,
c_index = c_index+1;
c(c_index,1) = c_int;
c(c_index,2) = c_freq;
index = index+1;
c_int = b(index,1);
c_freq = b(index,2);
end
23) thresh = calculate thresh-value of matrix c_index
24) L1 = quantized the thresh value
25) for len 1 to length_k
for breath 1 to breath_k
check, if L1(len,breath) ==1
then, L(len,breath) = threash value of 1st
value of matrix L
else
if L1(len,breath) = c_index
then, L(len,breath) = 255
else
L(len,breath) = calculate thresh value of L(L1(len,breath))
end
end
end
26) Img_edge = evaluate the edge of L matrix
27) Display Img_edge;
Algorithm for Segmentation, Object Detection and Object Extraction
1) Accept input image from user
2) Convert the input image into greyscale image
// Thresholding
3) I = input image
4) rmat = I(:,:,1);gmat = I(:,:,2);bmat = I(:,:,3); // divide each color channel matrix
5) set a threshold value of each color channel;
6) convert each channel into binary image
7) take sum of each channel
8) Display the sum channel as final binary channel

(J4R/ Volume 03 / Issue 08 / 003)
Segmentation Technique
//Complement the binary channel and fill the holes
9) Take complement of the binary channel
10) Fill each small holes in the complement binary image
11) Display the final complement channel as Segmented Image
Object Detection
12) Create morphological structure of grey image
13) Open the morphological image
14) Measure the area and eccentricity of grey image objects
15) Label each object using connectivity method
16) Status = measure area and eccentricity of the labelled image
17) Calculate area;
18) Calculate eccentricity;
//using feature analysis count objects
19) index_obj = Find the eccentricity of the objects
20) Make a list of all values of eccentricities
21) Display it;
22) for index 1 to length(index_obj)
h = draw rectangular box around the objects
set the color of the bounding boxes
end
23) if index>0
then, calculate the number of objects in the image and display it
end
Object Extraction
24) After complementing the sum of binary channel, fill the all small holes and label all the objects in the image
25) Find the maximum value of the labelling object
26) For j 1 to max value of label
Find the row and column value corresponding j value
len = max(row) - min(row) + 2;
breath = max(col) - min(col) + 2;
find, target = zeros([len breath]));
sy = min(col) - 1;
sx = min(row) - 1;
for i = 1:size(row,1),
x = row(i,1) - sx;
y = col(i,1) - sy;
target(x,y) = calculate the row and column value of each object
end
Display each image ;
end
Algorithm for Medical images
1) Accept input image from user
2) Convert the input image into greyscale image
3) Convert the input image into binary image
4) Display the binary image
5) Finding the edge of the input image by ‘canny edge detector’ method
6) Display the edge of the image
Thresholding and segmentation
7) Accepting a threshold value of grey image. Check if the value of grey image greater than the threshold value or not.
8) Take the complement of the threshold image.
9) Fill the holes in the complement image.
10) Display it as Segmented image
Object Detection
11) Create morphological structure of grey image
12) Open the morphological image

(J4R/ Volume 03 / Issue 08 / 003)
13) Measure the area and eccentricity of grey image objects
14) Label each object using connectivity method
15) Status = measure area and eccentricity of the labelled image
16) Calculate area;
17) Calculate eccentricity;
//using feature analysis count objects
18) index_obj = Find the eccentricity of the objects
19) Make a list of all values of eccentricities
20) Display it;
21) for index 1 to length(index_obj)
h = draw rectangular box around the objects
set the color of the bounding boxes
end
22) if index>0
then, calculate the number of objects in the image and display it
end
Object Extraction
23) After complementing the sum of binary channel, fill the all small holes and label all the objects in the image
24) Find the maximum value of the labelling object
25) For j 1 to max value of label
Find the row and column value corresponding j value
len = max(row) - min(row) + 2;
breath = max(col) - min(col) + 2;
find, target = zeros([len breath]));
sy = min(col) - 1;
sx = min(row) - 1;
for i = 1:size(row,1),
x = row(i,1) - sx;
y = col(i,1) - sy;
target(x,y) = calculate the row and column value of each object
end
Display each image ;
end
VI. CONCLUSIONS
In this paper, we have presented Image object extraction techniques. The general process of the Image object extraction has been
described. The Image object extraction techniques have been classified and discussed in detail.
REFERENCES
[1] F.Ciontault, D.Gurin, J-P.Guillemin, B.Chopinet, “Infield Triticum aestivum ear counting using color-texture image analysis”, New Zealand Journal of crop
and horticultural Science, pp.117-130,2008.
[2] Hakeem AejazAslam ,“A New Approach to Image Segmentation for Brain Tumor detection using Pillar K-means Algorithm”, IJARCCE, pp 1429-
1436,2013.
[3] C.Sriramakrishnan, “Performance Analysis of Advanced Image Segmentation Techniques”, International Journal of Computer Applications, pp 13-18,
2012.
[4] Khang Siang Tan ,“Color image segmentation using histogram thresholding – Fuzzy C-means hybrid approach”, ELSEVIER, pp 1-15, 2011.
[5] Alzate, C, “Image Segmentation using a Weighted Kernel PCA Approach to Spectral Clustering”, IEEE, CIISP 2007, pp 208-2013, 2007.
[6] Rafika Harrabi , “Color image segmentation using multi-level thresholding approach and data fusion techniques: application in the breast cancer cells
images ”, Springer, pp 1-12, 2012.
[7] DibyaJyoti Bora, “A Novel Approach Towards Clustering Based Image Segmentation”, IJESE, pp 6-10, 2014.
[8] Mei Yeen Choong “An Image Segmentation using Normalised Cuts in Multistage Approach” IJSSST, pp 10-16, 2016.
[9] Rajiv Kumar , “Image Segmentation using Discontinuity-Based Approach”, IJMIP, pp 72-78,2011.
[10] Amanpreet Kaur , “A Review Paper on Image Segmentation and its Various Techniques in Image Processing”, IJSR, pp 12-14, 2014.
[11] Muhammad Waseem Khan et al, “A Survey: Image Segmentation Techniques”, International Journal of Future Computer and Communication, pp 89-93,
2014.
[12] RohanKandwal, “Review: Existing Image Segmentation Techniques”, IJARCSSE, pp 153-156, 2014.
[13] M. Jogendra Kumar, “REVIEW ON IMAGE SEGMENTATION TECHNIQUES”, IJSRET, pp 992-997, 2014.

OBJECT DETECTION, EXTRACTION AND CLASSIFICATION USING IMAGE PROCESSING TECHNIQUE

More Related Content

What's hot (20)

Similar to OBJECT DETECTION, EXTRACTION AND CLASSIFICATION USING IMAGE PROCESSING TECHNIQUE (20)

More from Journal For Research (20)

Recently uploaded (20)

OBJECT DETECTION, EXTRACTION AND CLASSIFICATION USING IMAGE PROCESSING TECHNIQUE