Digital Camera Workflow for High Dynamic Range Images Using a Model of Retinal Processing

We propose a complete digital camera workflow to capture and render high dynamic range (HDR) static scenes, from RAW sensor data to an output- referred encoded image. In traditional digital camera processing, demosaicing is one of the first operations done after scene analysis. It is followed by rendering operations, such as color correction and tone mapping. Our approach is based on a model of retinal processing of the human visual system (HVS). In the HVS, rendering operations, including adaptation, are performed directly on the cone responses, which corresponds to a mosaic image. Our workflow conforms more closely to the retinal processing model, performing all rendering before demosaicing.. This reduces the complexity of the computation, as only one third of the pixels are processed. This is especially important as our tone mapping operator applies local and global tone corrections, which is usually needed to well render high dynamic scenes. Our algorithms efficiently process HDR images with different keys and different content