A NOBLE APPROACH TO DETECT SALIENT REGION FROM IMAGES

Abstract

Visual saliency is the perceptual quality that captures our attention and makes an object, person, or pixel stand out relative to its neighbors. The focus of this report is to design an algorithm for the automatic detection of visually salient regions in images, which is useful in applications such as adaptive content delivery, adaptive region-of-interest based image compression, image segmentation, object recognition, and content aware image resizing. In this work, an algorithm is designed to detect salient regions from images based on both global and local features and results are compared with several states of art methods given in literature. We found that our algorithm works best in term of precision, recall and f-measure. Also the computational complexity of our algorithm is very low so it can be used for real time applications.