Advances in Electrical and Computer Engineering Volume 21, Number 1, 2021 Fuzzy Contrast Enhancement System with Multiple Transform Domain Operations Tariq JAVID 1 , Muhammad ABID 2,3 1 Faculty of Engineering Sciences & Technology, Hamdard University, Karachi, Pakistan 2 Interdisciplinary Research Center, COMSATS University Islamabad, Wah Campus, Pakistan 3 Department of Mechanical Engineering, COMSATS University Islamabad, Wah Campus, Pakistan tariq.javid@hamdard.edu.pk 1 Abstract—Medical images provide an excellent way to identify the presence of diseases in a highly acceptable and correct manner. These images, however, due to the physical limitations of imaging instruments, have various types of artifacts. Techniques, such as those found in the computer- based image processing discipline are used as an alternative to the costly instrument to reduce or control these artifacts. A medical expert performs the complex task of disease identification and prognosis based on the visualized medical image data. The quality of medical images thus plays an important role to lower the chance of misdiagnosis and resulting in incorrect treatment. To meet the requirement of high-quality image data for the medical professional, in this research work, an innovative system is developed with the help of standard transform domain operations, data fusion, and fuzzy contrast enhancement system. Furthermore, the graphics processing unit and lookup table based technique are combined toward potential real-time implementation of the designed system. The proposed system can significantly improve the radiological contents inside medical image data to ease the task of the medical expert. Index Terms—Fourier transform, fuzzy systems, image enhancement, image fusion, wavelet transforms. I. INTRODUCTION Medical images are the representations of human body parts generated through complex medical imaging techniques. These acquired images are processed, enhanced, and analyzed through specialized hardware and software. Enhanced images are more useful to medical experts and ease the complex process of disease diagnosis. Figure 1 shows an example. In this example, the standard global histogram equalization is applied to the input low contrast X-ray image to generate the enhanced version [1]. There are numerous variants of histogram equalization in the literature [2-6]. These efforts have improved the standard global histogram equalization either through modification or by integration with other well-established image processing and analysis techniques. Fuzzy set theory was introduced to deal with the imprecise information [7-9]. The concept evolved gradually in the form of numerous successful applications in medical diagnosis [10-12]. The fuzzy rule-based system is an example of such an approach developed in this research area. A fuzzy rule-based contrast enhancement system is implemented in [13]. The application of such a system results in a better quality enhanced output image as compared to the standard histogram equalization (see Fig. 2 and Fig. 3). In Fig. 2, the output of the standard histogram equalization, that is, the middle image results in the improved contrast, however, the artifacts are more visible and affected the adjacent pixels. These artifacts are less obvious in the output of fuzzy contrast enhancement as in the left side image in Fig. 2, and similarly in the output of the Fig. 3. This work was supported by the Hamdard University under Grant no. HURC-16-048/2017. Many researchers have focused on developing innovative data-driven hybrid models and multi-model data fusion approaches to increase accuracy and precision. Figure 4 shows the schematic algorithms of hybrid models and multi- model data fusion approaches [14]. In Fig. 4(a), the models are applied to the input data in a series manner. In Fig. 4(b), the models are applied in parallel and all outputs are then fused to obtain the output. A statistical method based on K- NN estimation is proposed for data fusion in [15] that used input data statistics to improve the fusion task. Figure 1. X-ray and the enhanced image obtained by the application of the global histogram equalization Figure 2. From left, the input low-contrast image, the output histogram equalized image, and the output of fuzzy contrast enhancement [1] Figure 3. Gray scale input image and the enhanced output image through the application of fuzzy rule-based contrast enhancement system Transform domain representations provide an alternative to performing complex mathematical operations on images. Some mathematical operations in the transform domain are 83 1582-7445 © 2021 AECE Digital Object Identifier 10.4316/AECE.2021.01009