Local Features and Takagi-Sugeno Fuzzy Logic based Medical Image Segmentation

Abstract

This paper presents an improved region scalable fitting model that uses fuzzy weighted local features and active contour model for medical image segmentation. Local variance is used with local entropy to extract the regional information from the image which is then processed with the Takagi-Sugeno fuzzy system to compute weights. The use of regional descriptors enables this model to segment the inhomogeneous intensity images. The proposed objective function is minimized by using level set function. Performance evaluation of the proposed and existing model is achieved with the help of a Probability Rand Index, Global Consistency Error, the number of iterations and computation time taken. Extensive experiments on a series of real X-ray and MRI medical images shows the proposed technique offers better segmentation accuracy in lesser number of iterations and computation time.