[25]

Region-based Segmentation

ABCD rule +GLCM

Asymmetry, border, color, and differential structure. TDS = [(A × 1.3) + (B × 0.1) + (C × 0.5) + (D × 0.5)], Energy, Entropy, Contrast, Correlation, and Homogeneity.

BPNN

Due to NN based system, speed up the system performance.

A hybrid classification algorithm is needed.

90.45

[26]

-

GLCM + ABCD rule

ABCD and textural features are extracted.

BPNN

Due to the based system, it is secure and reliable.

The segmentation method is not properly maintained.

75.00%

[28]

-

DCT + DWT + SVD (Singular value decomposition)

Different coefficients are extracted which classify through SVD.

-

Due to mobile-based apps, it is used in mobile hospitals in rural remote areas with a lower computational time of 2.066 s.

Lack of segmentation and classification method, system performance is relatively poor.

80%

[33]

-

GoogleNet + AlexNet + VGGNet

The convolution layer-based approach generates features for classification.

-

Skin diseases are classified as Nevus, melanoma, Seborrheic Keratosis with lower computational time efficiency with the recall rate is 84.8%.

Segmentation and classification algorithm is not defined.

83.8

[34]

Otsu thresholding

GLCM + NGTDM

23 color and texture features are as Rmin, Gmin, Bmin, Rmax, Gmax, Bmax, Rmean, Gmean, Bmean, Hmean, Vmean, Cbmean, Crmean, Graymean, contrast, correlation, energy, and homogeneity whereas the coarseness, busyness, complexity, contrast and texture length.

SVM with quadratic kernel

Though images are taken from diverse sources system accuracy is satisfactory.

Few images are produced the same features which deteriorate the system performance adversely.

83%

[35]

-

DWT + PCA

Wavelets divide an image into 4 sub-band components which correspond to approximation, horizontal, vertical, and diagonal

respectively.

DT, Naive Bayes

This system is efficient on a small dataset and easy to use.

It is not adequate in larger datasets.

98.8%