SKIN CANCER LESION CLASSIFICATION USING LBP BASED HYBRID CLASSIFIER

richa sharma, Madan Lal

Abstract


Skin cancer is a most dangerous type of cancer found in humans. It is found in various types such as melanoma, basal cell carcinoma and squamous cell carcinoma. Among others, Melanoma is the most serious and dangerous cancer which leads from a simple skin mark to a tumour. Early detection of the type of skin cancer can helps in better cure. In this paper, a new method is proposed for the detection of type of skin cancer. The proposed method integrates the features of DRLBP (Dominant rotated local binary pattern) and MRELBP (median robust extended local binary pattern) methods for feature extraction of the skin cancer lesions. The support vector machine (SVM) classifier is then used to classify the images based on the calculated features. To evaluate the performance of the proposed method, skin cancer images containing 367 lesions are used from ISIC standard dataset from the results, it is analysed that the proposed method can extracts microstructure and macrostructure texture information. Furthermore, it is robust to the rotation variation. It is also observed that the proposed method gives better results than the other state of the art local binary pattern based feature extraction methods.

Keywords


skin cancer, texture recognition, Local binary pattern, texture classification.

Full Text:

PDF


DOI: https://doi.org/10.26483/ijarcs.v8i7.4469

Refbacks

  • There are currently no refbacks.




Copyright (c) 2017 International Journal of Advanced Research in Computer Science