AUTOMATIC DIAGNOSIS OF SKIN LESION SEGMENTATION USING TEXTURAL DISTINCTIVENESS

S Amanda, Elsa Nivitha
International Journal of Computer Technology & Applications   unpublished
Melanoma a potentially fatal form of cancer has many diagnostic approaches. But they still lacks in efficiency due to factors like complex visual patterns, streaks, presence of hair and cost of dermatological examining. This paper describes a method that differentiates a normal skin from that of a malignant type. With the acquired image, pre-processing is done by converting them to gray scale and dull razor algorithm is applied to remove fine hair, small glows and noises. Features are extracted
more » ... by acquiring the darker pixels found on the surface following which the radial distance between each dark spot is calculated. Segmentation is done using textural distinctiveness algorithm that calculates the local texture vectors by TD metrics and by computing the regional TD metrics the resultant map is acquired. Finally, classification of the segmented image is done with the use of support vector machine, where dataset is trained and their feature vectors are used to deduct generic equations, when a new dataset is tested in the system it predicts if the acquired image is melanoma or non-melanoma.
fatcat:uagie5abczaz7hunm4a36lrc6e