International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 03 Issue: 05 | May-2016 www.irjet.net p-ISSN: 2395-0072
© 2016, IRJET | Impact Factor value: 4.45 | ISO 9001:2008 Certified Journal | Page 3223
SEGMENTATION OF SKIN LESION FROM DIGITAL IMAGES USING
MORPHOLOGICAL FILTER
M.Yuvaraju
1
, D.Divya
2
, A.Poornima
3
1
Assistant Professor, Department of EEE, Anna University Regional Campus, Coimbatore, Tamil Nadu, India
rajaucbe@gmail.com
2
Assistant Professor, Department of EEE, Anna University Regional Campus, Coimbatore, Tamil Nadu, India
divyadevarajan@gmail.com
3
PG Scholar, Department of EEE, Anna University Regional Campus, Coimbatore, Tamilnadu, India
a.poorni69@gmail.com
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Skin cancer is the deadliest form of skin
disease. Its incidence has been rising at a rate of 3% per
year. In order to reduce the cost of screening, there is a
need for an automated melanoma screening system.
Segmentation is significant to detect skin lesion from
images. In the proposed method, a novel texture based
skin lesion segmentation algorithm is used and
probabilistic neural network is used to classify the stages
of skin cancer. The feature of the image is extracted by
using GLCM algorithm and its features gives better
classification with probabilistic neural network. The five
different skin lesion is commonly grouped into Basal Cell
Carcinoma (BCC), Actinic Keratosis (AK), Squamous Cell
Carcinoma (SCC), Melanocytic nevus/mole (MC),
Seborrhoeic Keratosis (SK).The system will be used to
classify the queried images automatically to choose the
stages of abnormality. The morphological filter
segmentation is used to detect the skin cancer. The
proposed system has higher accuracy, sensitivity,
specificity, segmentation compared to other systems.
Key Words: Skin cancer, Grey level co-occurrence
matrix, Probabilistic Neural Network, Segmentation.
1. INTRODUCTION
Skin is commonly used primal in image processing and
the applications range from face tracking to signal
analysis for different human interactions. Generally
skin cancers are the most common prevalent form of
cancers in human beings. Surprisingly it is also a
deadly type of cancer. Many of the skin cancers are
curable at early stages. Also with the technology
advancements, early detection is possible [1]. The
American cancer society estimates that more than
70,000 new skin cancers are diagnosed every year in
the United States alone. The skin cancers can be
classified into melanoma and non-melanoma.
Melanoma is the most deadly form and are predictable
76,690 people being diagnosed with melanoma and
9480 people quiet of melanoma in the United States. In
the United States the life time hazard of receiving
melanoma is 1 in 49 [2].
Melanoma reasons for approximately 75% of deaths
related with skin cancer. It is a malignant tumor of the
melanocytic and generally happens on the trunk or
lower extremities. Recent trends have stated that
melanoma can be less dangerous if detected at a
premature stage. I.e. if detected at stage I the survival
rate of the effected person increase to 96% [3]. Due to
the increase in the incidence rate of melanoma,
researchers are more concerned about proposing such
automated systems that diagnose skin lesion correctly.
Also it has been found that in order it detect Melanoma
at an early stage screening is very much valuable [4]
but the cost of screening melanoma is too high. So to
reduce the screening cost the automated algorithms
have been proposed to automatically screen
melanoma.
A digital dermoscope acquires images that contribute
to early screening of melanoma [5] and all automated
systems use dermoscopic images. Dermoscope is a
device that is used to capture images of lesion by the
dermatologists. It also magnifies the image and acts as
a filter. With dermoscopy it becomes difficult to
differeniate malignant and benign lesion and in such
case a detailed analysis is needed to be done [6].
Recent work with automated melanoma screening