Human Vision Based Framework for the Extraction of Salient Regions

Mechanisms which are not fully understood today, exist in human visual system that reduces the bulk of input sensory data and makes possible for visual system to analyze a scene in real time. The reduction of sensory input is based on saliency map i.e. a topographical map of salient regions, of the scene. Research work presented in this book has investigated the effect of blur on human visual attention. Hypothesis has been investigated that the sharp regions tend to capture attention irrespective of intensity, color or contrast. The second objective of this research was to find such algorithm of saliency detection that performs adequately in predicting human fixation for stimuli containing blur and sharp regions. Framework is proposed for the extraction of salient regions specifically for stimuli containing blur and sharp regions. Data gathered from visual experiment is used as the ground truth for identifying salient regions for 122 images (specifically built database) and compared with the salient regions extracted by new approach. Proposed approach for extraction of salient regions out performs state-of-the-art saliency detection methods in terms of precision and recall.

Autorentext

Rizwan Ahmed Khan is a PhD scholar at LIRIS, Université Lyon 1, France. He has completed his Masters in "Colors in Informatics and Media Technology" from Université Jean Monnet, France and Gjøvik University College, Norway. His current research interests include computer vision, image processing and pattern recognition.