Performance improvement of MF-DFA on feature extraction of skin lesion images

Abstract

In this paper, we propose an improved algorithm based on the original two-dimensional (2D) multifractal detrended fluctuation analysis (2D MF-DFA) that involves increasing the number of cumulative summations in the computational steps of 2D MF-DFA. The proposed method aims to modify the distribution of the generalized Hurst exponent to ensure that skin lesion image features are extracted based on enhanced multifractal features. We calculate the generalized Hurst exponent using 0, 1, or 2 cumulative summation processes. A support vector machine (SVM) is adopted to examine the classification performance under these three conditions. Computation shows that the process involving two cumulative summations achieves an accuracy, sensitivity, and specificity of [Formula: see text], [Formula: see text], and [Formula: see text], respectively, which indicates that its performance is much better than with 0 and 1 cumulative summations.

Research on image features extraction based on machine learning algorithms

“…With network technology and intelligent search engines, innovative artificial intelligence technologies have brought a boom in visual image data for new Internet of Things (IoT) technologies, etc. Artificial intelligence technologies represented by machine learning and deep learning have also greatly promoted the cognitive ability of machines to recognize image features and image recognition [5]. However, the explosive growth of the distribution of real physical visual data presents significant characteristics of high complexity, dynamism and data privacy.…”