Session 02 | Single Image Dehazing Method скачать в хорошем качестве

Session 02 | Single Image Dehazing Method

IEEE Computer Society Bangalore Chapter - BITS APPCAIR - Joint Webinar Series 2020 | Session 02 Single Image Dehazing Method by Prof. Bhabatosh Chanda About the topic: Quality of captured image not only depends on the technical specification of the optical sensor or camera, but also on the intensity and distribution of light over the scene as well as the clarity of the scene. During unfavourable condition, such as rain, fog or haze, the quality of captured image degrades significantly. Mathematical model describing haze or fog leads to ill-posed problem. So to solve the problem or, in other words, to restore haze-free image from hazy ones becomes a challenge. Available methods either use multiple images or use additional knowledge about the scene. Image dehazing from a single image is even more difficult. Present talk discusses the problem statement, brief overview of existing methods, and then two approaches to address the problem. First approach considers imaging model, relaxes it to accommodate both day-time and night-time imaging. Treatment provides an insight to cause and remedy of the dehazing problem. The second approach emphasizes on the parameters such attenuation and air-light responsible for hazing effect, and propose a data drives solution using machine learning techniques.