Open Access
A modified fractal texture image analysis based on grayscale morphology for multi-model views in MR Brain
Author(s) -
R Usha,
K. Perumal
Publication year - 2021
Publication title -
indonesian journal of electrical engineering and computer science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.241
H-Index - 17
eISSN - 2502-4760
pISSN - 2502-4752
DOI - 10.11591/ijeecs.v21.i1.pp154-163
Subject(s) - grayscale , artificial intelligence , pattern recognition (psychology) , fractal dimension , fractal analysis , image texture , fractal , segmentation , box counting , computer science , texture (cosmology) , image segmentation , computer vision , feature extraction , transformation (genetics) , mathematical morphology , binary image , image (mathematics) , mathematics , image processing , mathematical analysis , biochemistry , chemistry , gene
This paper presents a modified fractal texture feature analysis with the use of grayscale image morphology for automatic image classification of different views in MR brain images into normal and abnormal. This main contribution of this approach is a reduction of the total number of a threshold value, and the number of image decomposition, in which only the number of extract threshold value two or three are enough for tumor region extraction - compared to four or more is required in the previous method of SFTA (segmentation based fractal texture analysis). This is achieved by pre-processing of hierarchical transformation technique (HTT), which make use of morphological image transformations with the desired structural element. From this decomposed images, mean, area, fractal dimension and selective shape features are extracted and fed into KNN and ensemble bagged tree classifiers. Finally, some of the post-processing is handled for tumor region extraction and tumor cells computation. It is found that this proposed approach has superior results in the segmentation of diseased tissue from normal tissue and the prediction of image classes in terms of accuracy with the less number of threshold extraction and image decomposition rather than the SFTA algorithm.