Image and Video Processing based on Kernel Representations

Computer vision is a field of artificial intelligence which aims to electronically perceive and understand dynamics from high-dimensional data obtained from the real world in order to produce numerical or symbolic information. Due to the fast spread of digital cameras and the impressive technical enhancements performed during the last decade, the development of new computer vision algorithms has become a research topic of great importance. In this work, different kernel-based feature representation frameworks are proposed. Our main goal is to properly reveal the most relevant information from high dimensional data for enhancing the performance of two different computer vision algorithms: image and video segmentation. To this, we propose to use Multiple Kernel Representations (MKR) to incorporate multiple image features, such as: color representations, pixel spatial information and optical flow-based information.