Malignant melanoma is one of the most aggressive forms of skin cancer which must be necessary to be diagnosed at the initial stage for effective treatment. Melanoma affects the patient life even it can become a reason of death if its diagnosis is not accomplished on time. Through a rough pigment network and some suspicious signs can be helpful for diagnosis the melanoma from dermoscopic images. According to the clinical studies, for dermatologists, it is quite difficult to identify these signs at the initial stage of melanoma. So, it is important to propose an automated system which can efficiently be identified and differentiate between benign and malignant melanoma. The main focus of this research article is to improve the skin lesion segmentation from low contrast and under/ over segmented dermoscopic images through fusing the region based active contour method with JSEG method. The proposed fused segmentation technique gain 95.3% accuracy and through our proposed feature vector the Gaussian classifier achieved the promising results as sensitivity 97.7%, specificity 96.7%, and accuracy 97.5% with handling the special dermoscopic image cases which are comparatively much better than numerous exiting techniques.