|
| Reference | Approach | Disadvantages |
|
| Stoecker et al. [15] | Gray-level co-occurrence matrix for texture feature | The presence of artifacts like shining areas and shadows caused by light makes the process of segmentation of skin lesion images more complicated. |
|
| Stoecker and Scharcanski [22] | Four different algorithms | To identify the region of nuclei which used the intensity and size of nuclei as a parameter |
|
| Sonali and Kamat [23] | Combined thresholding with fuzzy C-means | It may not perform well over images with huge variations in skin colors |
|
| Manju Bharathi and Sarswati [24] | NC ratio analysis for automatic segmentation of cells | Performance degrades over lesions of varying sizes and shapes |
|
| Jeniva and Santhi [25] | Learning model of natural skin texture and cancer textures | A lot of difference between specific kinds of cancer and the surrounding area of skin |
|
| Kumar et al. [26] | Local region recursive segmentation, K-means clustering, and local double ellipse descriptor | It may not perform well over images with huge variations in skin colors |
|
| Abbadi and Miry [27] | Thresholding and Wiener filter | Low lesion-to-skin gradient, depigmentation, multiple tumor regions |
|
| Lu et al. [28] | Mean shift, local region recursive segmentation, and local double ellipse descriptor | This method is computationally complex |
|
| Baral, Gonnade, and Verma [29] | Neuro-fuzzy model and some other features | Complex thresholding approaches |
|
