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

Image tone mapping approach using essentially non-oscillatory bi-quadratic interpolations combined with a weighting coefficients strategy

“…Simulation results are compared to competitive methods : (i) SEP [10] with β = 0.3, γ = 0.7, J=2, upd=0, b = e ; (ii) Li TMO [11] with Haar multiscale ; (iii) Fattal [9] using RBW method with α = 0.8, β = 0.3, γ = 0.8, J=2, upd=0 ; (iv) Drago [2] , Reinhard [3] , Ward [4] , Durand [5] , Tumblin [6] , Schlick [7] with the default parameters as given in the HDR Toolbox ; and (v) Duan [8] using β = 0.5. For the proposed method, denoted "Proposed", simulations are conducted with the following parameters : ε = 10 −9 , J = 2, γ = 0.6, ξ = 0.1, σ i deduced from equation (15).…”

“…Its strength points are inherent not only to the non-separable multiresolution decomposition but also to the ENO strategy. It suffices to compare its performance with that of separable multiresolution ENO strategy developed in [10].…”

“…Anturium Bottle Office Church Memorial Light Drago [2] 0.874 0.801 0.801 0.814 0.800 0.800 Reinhard [3] 0.778 0.807 0.826 0.789 0.792 0.794 Ward [4] 0.806 0.783 0.775 0.817 0.793 0.789 Durand [5] 0.811 0.892 0.825 0.929 0.814 0.800 Tumblin [6] 0.715 0.713 0.735 0.675 0.761 0.750 Schlick [7] 0.770 0.835 0.926 0.970 0.788 0.780 Duan G [8] 0.964 0.916 0.955 0.986 0.935 0.969 Duan L [8] 0.910 0.924 0.927 0.906 0.839 0.924 Fattal [9] 0.889 0.928 0.928 0.889 0.927 0.971 SEP [10] 0.888 0.931 0.928 0.885 0.903 0.960 Li [11] 0…”

“…This approach was then exploited to map an HDR image into an LDR image. In [10], the authors proposed non-linear separable multiresolution families using data dependent interpolation where a set of operators have been derived. These latter have the ability to consider in their mathematical models the image singularity points.…”

Revisited histogram equalization as HDR images tone mapping operators

“…Recently, significant number of Tone Mapping Operators (TMOs) converting the HDR images to LDR images have been proposed using different strategies. Among the developed methods, some of them were designed with respect to the Human Visual System (HVS) [1], another were based on the frequency domain [2]- [4], segmentation operators were also designed to combine multiple functions to different segments [13], [14]. However, tone mapped images suffer from data loss because of the reduction process into low dynamic range, affecting the quality of the tone mapped images.…”