Performance evaluation of high dynamic range image tone mapping operators based on separable non-linear multiresolution families

Abstract

This paper addresses the conversion problem of High Dynamic Range (HDR) images into Low Dynamic Range (LDR) images. In this objective, separable non-linear multiresolution approaches are exploited as Image Tone Mapping Operators (TMOs). They are related on: (i) Essentially Non-Oscillatory (ENO) interpolation strategy developed by Harten namely Point-Value (PV) multiresolution family and Cell-Average (CA) multiresolution family; and (ii) Power-P multiresolution family introduced by Amat. These approaches have the advantage to take into account the singularities, such as edge points of the image, in the mathematical model thus preserving the structural information of the HDR images. Moreover the Gibbs phenomenon, harmful in tone mapped images, is avoided. The quality assessment of the tone mapped images is measured according to the TMQI metric. Simulation results show that the proposed TMOs provide good results compared to traditional TMO strategies.