Abstract:
Low-field MRI scanners do not offer sufficient image contrast. Hence, offline algorithms for improving image contrast are often needed. Even though modified versions of Histogram Equalisation (HE) are extensively used on panoramic images, they have serious limitations. Most of such modified algorithms have multiple operational parameters which need to be tuned manually. Parameter-free modifications lag in terms of illumination-preserving features. To address these issues, a novel formulation of Parameter-free Fuzzy Histogram Equalisation (PFHE) algorithm with good illumination-preserving characteristics, dedicated for contrast enhancement of MRI is introduced in this paper. In PFHE, a Homogeneity Fuzzy Sub-set (HFS) and its fuzzy complement, termed as Texture Fuzzy Sub-set (TFS) are computed based on the fuzzy similarity of the pixels in the input image with their eight-connected neighbours. Following this, an approximate output is estimated by applying a transformation similar to the histogram equalisation on the Fuzzy Textural Histogram (FTH) derived from TFS. The final output is computed as a nonlinear combination of the approximate output and the input image. The fuzzy weighting vectors used in the nonlinear combination are derived from the HFS. Both Qualitative and quantitative evaluations reveal that the PFHE is superior to Bi-Histogram Equalisation (BHE), Weighted Threshold Histogram Equalisation (WTHE), Contrast Limited Adaptive Histogram Equalisation (CLAHE), Non-parametric Modified Histogram Equalisation (NMHE), Exposure-based Sub-Image Histogram Equalisation (ESIHE), Median–Mean Based Sub-Image-Clipped Histogram Equalisation (MMSICHE) and Dominant Orientation-based Texture Histogram Equalisation (DOTHE), in terms of ability to preserve diagnostically significant features in the MR image.