Digital cameras can only capture a limited range of real-world scenes' luminance, producing images with saturated pixels. Existing single image high dynamic range (HDR) reconstruction methods attempt to expand the range of luminance, but are not able to hallucinate plausible textures, producing results with artifacts in the saturated areas. In this paper, we present a novel learning-based approach to reconstruct an HDR image by recovering the saturated pixels of an input LDR image in a visually pleasing way. Previous deep learning-based methods apply the same convolutional filters on wellexposed and saturated pixels, creating ambiguity during training and leading to checkerboard and halo artifacts. To overcome this problem, we propose a feature masking mechanism that reduces the contribution of the features from the saturated areas. Moreover, we adapt the VGG-based perceptual loss function to our application to be able to synthesize visually pleasing textures. Since the number of HDR images for training is limited, we propose to train our system in two stages. Specifically, we first train our system on a large number of images for image inpainting task and then fine-tune it on HDR reconstruction. Since most of the HDR examples contain smooth regions that are simple to reconstruct, we propose a sampling strategy to select challenging training patches during the HDR fine-tuning stage. We demonstrate through experimental results that our approach can reconstruct visually pleasing HDR results, better than the current state of the art on a wide range of scenes.

The quality of the images obtained from mobile cameras has been an important feature for modern smartphones. The camera Image Signal Processing (ISP) is a significant procedure when generating high-quality images. However, the existing algorithms in the ISP pipeline need to be tuned according to the physical resources of the image capture, limiting the final image quality. This work aims at replacing the camera ISP pipeline with a deep learning model that can better generalize the procedure. A Deep Neural Network based on the UNet architecture was employed to process RAW images into RGB. Pre-processing stages were applied, and some resources for training were added incrementally. The results demonstrated that the test images were obtained efficiently, indicating that the replacement of traditional algorithms by deep models is indeed a promising path.