Table 1.
Determining asymmetry of skin lesions in literature.
Table 2.
Characterizing color variegation in literature.
Table 3.
Measuring skin lesion diameter in literature.
Fig 1.
Proposed skin lesion segmentation approach.
Fig 2.
The original color image and its extracted lesion.
Fig 3.
(a) Extracted skin lesion and its four halves along the vertical and horizontal axes (b) Keypoints (in red) of each half (c) Matching keypoints between each half (top figure: Left and right halves, bottom figure: Top and bottom halves).
Fig 4.
Histograms of horizontal and vertical projections.
Fig 5.
(a) 3D scatter plot of skin lesion image pixels in CIELab color space (b) Skin lesion color palette showing the 7 dominant colors in the image (the first color which represents the pure black color will be omitted from our color variegation measurement).
Table 4.
CIELab melanoma suspicious color values.
Fig 6.
(a) Skin lesion image with a ruler used to spatially calibrate our test images and deduce the pixels/mm value (b) Skin lesions’ Feret’s diameter (in pixels and millimeters).
Fig 7.
(a) Skin lesion images segmented using Otsu and Otsu-II methods, along with their corresponding groundtruth (b) Skin lesion images segmented using our proposed segmentation approach and U-Net, along with their corresponding groundtruth.
Table 5.
Dice similarity of Otsu related methods on three datasets.
Fig 8.
Samples of symmetric and asymmetric images labeled by the dermatologist.
Fig 9.
Skin lesion asymmetry, color variegation, and diameter features extracted using the proposed approach.