Uncertainty Minimization in the Human Visual Response Using an Interval Type-2 Fuzzy Set and Its Application to Rice Leaf Image Enhancement

The paper proposes a novel technique for uncertainty minimization in human visual response (HVR) for image contrast enhancement. The HVR divides an image into three regions: De Vries-Rose region (DVR), Weber region (WR), and Saturation region (SATR). The detection of a pixel belongs to which region depends on its immediate background and its incremental threshold. The uncertainties occur in HVR because of different nonlinearities in incremental thresholds in the three regions. To manage uncertainties, the proposed method divides each region into two zones: high-uncertainty region (HUR) and low-uncertainty (LUR) region. An interval type-2 fuzzy set (IT2FS) is then constructed based on these zones to represent the uncertainties. An image is enhanced by minimizing the uncertainties. The minimization problem for enhancement is solved using a hybrid leader-based optimization (HLBO) technique. We choose rice leaf images to test the efficiency of the proposed enhancement method. The different lighting conditions and the lack of high-definition captured devices make the rice leaf image patterns highly uncertain. Proper enhancement of a rice leaf image is necessary for the classification of diseases. The proposed method is compared to state-of-the-art enhancement methods on the standard rice leaf image dataset. It is found that the performance is quite satisfactory compared to the state-of-the-art conventional and other fuzzy-based methods.