HDR (High Dynamic Range) imaging techniques offer photographers the ability to capture the full range of luminance in real-world scenes, overcoming the limitations of capture and display devices. One popular method for creating HDR images is the multiple exposures technique (MET), which involves capturing multiple exposures with regular digital cameras and combining them later to generate an HDR image.
In this thesis, we focus on a specific method called Compressed Exposure Sequences (CES) that aims to consolidate all the information from a bracketed sequence into a single JPEG file. Typically, the main image displayed by a standard image viewer is selected from the middle exposure, although alternative exposures or tone-mapped HDR images can also be utilized. If needed, the original exposures can be reconstructed from this single JPEG file, enabling their use in a standard HDR workflow.
At the outset, our initial proposal revolved around a storage method that involved capturing the difference between each exposure and the reference. These differences were then compactly stored within the metadata section of the same file. However, upon a comprehensive assessment of the drawbacks associated with this approach, it became evident that opting for the direct compression of each exposure, as opposed to relying on differences, yielded better outcomes in terms of both file size and quality metrics. This realization prompted us to seek an alternative course of action.
Hence, our newly suggested approach centers on a patch-based procedure. This procedure entails the storage of patches pertaining to under-exposed, over-exposed, and motion-affected instances. Meanwhile, other patches are reconfigured through the implementation of the camera response function, thus minimizing potential data loss. Moreover, our endeavor extends beyond this by devising a residual learning model. This model serves to augment the reconstructed exposures generated through the patch-based method, effectively surpassing the limitations inherent in the patch-wise technique.