Techniques (Reference number is mentioned)

Segmentation

Feature extraction

Features

Classification

Advantages

Limitations

Accuracy

[14]

Color segmentations

DWT + PCA

Features are extracted from DWT which are reduced for better accuracy.

kNN

It requires less computational time and memory.

DWT requires huge capacity and is computationally more expensive.

97.5%

[17]

Normalized Otsu thresholding

GLCM+ PCA

Mean, standard deviation, variance, entropy, contrast/inertia, homogeneity, energy, correlation, area, perimeter, diameter, asymmetry index, circularity index, fractal dimension, compactness index

DLNN (Deep learning NN), SVM-Adaboost

This is a CAD system that runs with lower computational time with higher accuracy.

Hybrid segmentation is needed to enhance system performance.

93%

[18]

-

GLCM + LBP

Energy, entropy, contrast, homogeneity, and LBP array features.

SVM

System performances are computed both qualitatively and quantitatively.

The segmentation algorithm is undefined and to boost up system performance NN based classification is required.

90.32

[19]

Grab Cut algorithm

Histogram + ABCD rule

Features are the area of the lesion, perimeter of a lesion, eccentricity, mean, standard deviation, L1 norm, L2 norm

angle of lesion, major and minor axis of the lesion from the segmented image.

SVM

It is easy to access and use due to its Smartphone embedded applications.

An improper segmentation algorithm is used which degrades system performance.

-

[22]

Threshold-based Adaptive Snake (AS) approach

ABCD rule + Epiluminescence microscopy (ELM) criteria algorithm

Features extracted in ABCD rule with ELM criteria.

GA + SVM with Radial Basis Function

(RBF)

GA reduces the dimensions and also defines the most discriminating subsets of features to boost system

performance.

An efficient and reliable segmentation algorithm is required.

88%

[24]

Otsu thresholding

GLCM + ABCD rule + PCA

GLCM features are Energy, correlation, homogeneity, and contrast features. The best 5 features with maximum efficiency as follows: TDS, mean, standard deviation,

energy, and contrast respectively.

SVM

The computational complexity is relatively lower than others.

Hybrid segmentation and NN-based

classification are needed.

92.10%