A selective approach for tone mapping high dynamic range content

Alessandro Artusi, Despina Michael, Benjamin Roch, Yiorgos Chrysanthou, Alan Chalmers
2011 SIGGRAPH Asia 2011 Posters on - SA '11  
Problem Large resolution HDR textures/images are becoming common in computer graphics applications; but it is still a prohibitive task to perform on them a filtering step such as local tone mapping in real-time. Challenges Real-time performances for very large HDR textures/images can be achieved. Furthermore, this framework can be used to identify regions of the input HDR image where different TMOs or different settings of the same TMO can be applied to better preserve contrast and details. Our
more » ... st and details. Our Approach • Observation: Knowledge of where the strong contrast is located may be exploited to limit the use of a computational expensive local luminance adaptation computation only on these regions. • We propose a simple framework, Figure 1 , that allows to save high amount of computation time that may be re-addressed to other steps of the rendering pipeline. Observations • 1 st Observation: Gradient Magnitude is a good feature of contrast. • We would like to identify the areas that contribute to the luminance adaptation computation. Solution • Edge detector (Sobel Filter) applied on the logarithmic domain. • Thresholding allows to identify the lowest gradient transition to keep details in low and high contrast regions • 2 nd Observation: Saliency map applied on HDR images reveals that the high level of luminance areas are the one most salient. • On the other hand the saliency map is failing in identify the transition areas between bright and dark regions. • Future work needs to investigate how to exploit temporal coherence between multiple frames to enable at least significant parts of the importance areas map to be reused. • With respects to interactive local adjustment tool future work can exploit the use of the selective approach as a tool where the user interactively can select different areas and apply different local contrast adjustment
doi:10.1145/2073304.2073360 fatcat:mzskmlzlqbcnnkw6idbh7kbrxq