Abstract:
Visualization on low dynamic range (LDR) output devices like movie screens or standard displays requires a computationally fast tone mapping operator (TMO) that can quickly adapt to a wide range of HDR content.
Existing TMOs can tone-map a limited number of HDR content and require extensive parameter tuning to produce the best subjective-quality tone-mapped output. This paper proposes a fast, parameter-free, scene-adaptable deep tone mapping operator (DeepTMO) that produces high-resolution and high-subjective tone mapped output.
DeepTMO, based on conditional generative adversarial network (cGAN), learns to adapt to vast scenic-content (e.g., outdoor, indoor, human, structures, etc.) and HDR-related scene-specific challenges like contrast and brightness while preserving fine-grained details.
We investigate four Generator-Discriminator architectural designs to address blurring, tiling patterns, and saturation artifacts in HDR deep-learning frameworks. We choose a multi-scale model after exploring scales, loss-functions, and normalization layers in cGAN.
We train our network using Tone Mapping Image Quality Index (TMQI) to take advantage of the large amount of unlabeled HDR data. Our DeepTMO produces high-resolution, high-quality images across a wide range of real-world scenes. Finally, a pair-wise subjective study confirms our method’s versatility by assessing our results’ quality.
Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.
Did you like this final year project?
To download this project Code with thesis report and project training... Click Here
