Poster

Poster

Low-light video footage and ground truth collection

Ideal lighting conditions are not always on hand when capturing scenes. This is especially true when filming the natural environment, which can result in low-light footage. Consequently, sequences can suffer spatial temporal incoherent noise, flicker and blurring of moving objects.  Furthermore, scenes are complicated, and objects of interest may not be contained within the focal point of interest, which further degrades the quality of the scene. Manual correction using experts is both expensive and time consuming and cannot be easily scaled up. This poster will give an overview of our novel technique utilising deep learning architectures to resolve many of these issues. We will also highlight some of its limitations, including datasets, and recommendations to overcome these.

Justin Worsey

Bristol

Biography

−