IMAGE TONE MAPPING APPROACH USING ESSENTIALLY NON-OSCILLATORY BI-QUADRATIC INTERPOLATIONS COMBINED WITH A WEIGHTING COEFFICIENTS STRATEGY

Ba Chien Thai , Anissa Mokraoui (1) , Basarab Matei (2) , Mahdia Tabal (1)

1 L2TI - Laboratoire de Traitement et Transport de l'Information
2 LAGA - Laboratoire Analyse, Géométrie et Applications

This paper proposes a Tone Mapping Operator (TMO) to convert High Dynamic Range (HDR) images into Low Dynamic Range (LDR) images. It is related on two consecutive stages. The first one decomposes the HDR image on several resolution levels according to a non-separable multiresolution approach using Essentially Non-Oscillatory (ENO) strategy where adapted bi-quadratic interpolations are exploited. The underlying idea of this decomposition is to better represent the details of the complex HDR image preserving as much as possible the HDR image quality. In the second stage, a weighting coefficients process is deployed to reduce judiciously the dynamic range of each subband's coefficient. The weights are deduced from Gaussian filtering operations combined with a monotonic decreasing function where its parameters depend on the statistical properties of each HDR subband, neighborhood and resolution level. Simulation results show that the proposed image TMO has the ability to better represent the details compared to other operators.

Domaines: Sciences de l'ingénieur [physics]; Traitement du signal et de l'image [eess.SP]

Ba Chien Thai, Anissa Mokraoui, Basarab Matei, Mahdia Tabal. IMAGE TONE MAPPING APPROACH USING ESSENTIALLY NON-OSCILLATORY BI-QUADRATIC INTERPOLATIONS COMBINED WITH A WEIGHTING COEFFICIENTS STRATEGY. International Symposium on Signal Processing and Information Technology, Dec 2017, Bilbao, Spain. ⟨10.1109/ISSPIT.2017.8388688⟩. ⟨hal-02428266⟩