Adaptive Windowing for Optimal Visualization of Medical Images Based on a Structural Fidelity Measure

Medical imaging devices often capture the raw data with high precision, producing high dynamic range (HDR) images. To visualize HDR images on regular displays, there has been an increasing number of tone mapping algorithms developed in recent years that convert HDR to low dynamic range (LDR) images. To visualize HDR medical images, a so-called “windowing” procedure is typically employed by which the structural details within the intensity region of interest is mapped to the dynamic range of regular displays. Linear mapping is the most straightforward windowing operator, but may not be the optimal mapping function in terms of structure preserving. Here we propose a framework to adaptively find the optimal windowing function for different images. Specifically, a recently developed structural fidelity measure for tone mapped images is employed to adaptively optimize the windowing function, so as to achieve the best structural fidelity with respect to the original HDR image. Experiments demonstrate the promising performance of the proposed adaptive windowing method.