Minimum intensity error intuitionistic fuzzy contrast enhancement transformation for magnetic resonance images

Abstract

Background: Low contrast in magnetic resonance (MR) images adversely affects the performance of software tools used for its automated analysis. Compared to the traditional crisp transformation functions often used for enhancing the MR images, it is easy for the users to customize the fuzzy contrast transformation functions in order to selectively improve the contrast among the grey levels within any desired region of the dynamic range. However, computationally fast fuzzy-based image contrast enhancement techniques with good ability to preserve the mean brightness level are rare.Objectives: To resolve these issues, a computationally fast algorithm with good ability to preserve the mean brightness level, named minimum intensity error intuitionistic fuzzy contrast enhancement transformation (MIEIFCET) is proposed in this paper. Methods: In the MIEIFCET, the fuzzy set obtained from the input image by normalizing the pixel values in it, is represented as an intuitionistic fuzzy set (IFS) with the help of the Atanassov intuitionistic fuzzy generator. The contrast among the membership values in the IFS is improved by stretching them to the extremes by applying a customized fuzzy decision rule. The contrast-enhanced image is obtained by expanding the stretched membership values in the IFS to the full possible dynamic range. Results: On 100 MR images, the coefficient of variation (COV) of the output images of brightness-preserving dynamic fuzzy histogram equalization (BPDFHE), fuzzy contrast intensification (FCI), fuzzy enhancement for low-exposure images (FELI), fuzzy theory-based adaptive image enhancement (FTAIE), optimum fuzzy system for image enhancement (OFSIE), parameter-free fuzzy histogram equalisation (PFHE), type-2 fuzzy contrast enhancement (Type-2 FCE), fuzzy logic-based image enhancement (FLIE), contrast enhancement based on type‑2 intuitionistic fuzzy set (Type 2 IFS FCE), and MIEIFCET are 0.9023 ± 0.0751, 1.1219 ± 0.2526, 0.8992 ± 0.1263, 0.7236 ± 0.1644, 0.5315 ± 0.0445, 1.0664 ± 0.1517, 0.9199 ± 0.2441, 1.1895 ± 0.1863, 0.9331 ± 0.2312, and 1.2946 ± 0.1537, respectively. The computational time (sec) of BPDFHE, FCI, FELI, FTAIE, OFSIE, PFHE, Type 2 FCE, FLIE, Type 2 IFS FCE, and MIEIFCET are 0.1272 ± 0.1976, 0.0464 ± 0.0147, 0.7249 ± 0.2948, 0.0357 ± 0.0235, 0.3136 ± 0.1465, 0.3136 ± 0.1465, 0.0333 ± 0.0238, 0.0855 ± 0.0581, 0.0443 ± 0.0279 and 0.0169 ± 0.0155, respectively. Conclusion: Objective results in terms of highest COV and lowest computational time prove that the MIEIFCET has the ability to enhance the contrast of MR images without causing any significant change in the mean brightness level and it is computationally fast.