Techniques (Reference number is mentioned) | Segmentation | Characteristics | Feature extraction | Classification | Advantages | Limitations | Accuracy |
[15] | K-means clustering + Otsu method | Images are clustered with the threshold value. | GLCM | BPNN | An NN-based system that is in a real-time embedded manner. | Computational complexity is high due to it works in real time environment. | 95.83% |
[21] | Otsu Thresholding method + Edge-based Morphological operations | Edge-based segmentation with threshold values. | 2-D Fast Fourier Transform, 2-D Discrete Cosine Transform, Complexity Feature Set, and Color Feature Set | SVM | An enriched filtering algorithm helps all other stages to perform efficiently. | Well defined feature extraction algorithm is required to enhance accuracy. | 93.50% |
[27] | The active contour method + Marker control watershed algorithm. | The output of the active contour is used as an input to the marker control watershed algorithm | GLCM | SVM | A CAD system enhances the detection and classification and reduces the time latency. | An NN-based classification approach is needed. | 94% |
[30] | Uniform distribution based segmentation + Active contour method | Four metrics: DICE, Jaccard Index, Jaccard Difference, and Diameter are calculated for each segmented image. | Color features | SVM | It is a robust skin disease detection system with value. | Reliable and low complexity based feature extraction method is required. | 97.5% |
[31] | Edge-based segmentation + ACM | The active contours method works with sharp edges. | Sobel Operator | BPNN | An NN-based method that successfully detects several diseases. | An efficient hybrid feature extraction algorithm is needed. | 90% |
[32] | Otsu thresholding method + Morphological operations: dilation, erosion | A morphological operation performs on pixels. | ABCD rule | - | Almost all of the melanoma images are correctly identified using morphological operations. | Without a trained neural network, diseases are not classified properly. | - |
[37] | ACO + GA algorithm | Initialize GA and ACO parameters. Segmentation accuracy is 94%. | GLCM | Transductive Support Vector Machine (TSVM) | Better segmentation accuracy gives high performance. At first, this system is predicted as 24 diseases with fitness function. | Hybrid feature extraction is required. | 95% |