Automatic Region of Interest Detection in Natural Images

Identifying the Region or Object of Interest in a natural scene is a complex task because the content of natural images consists of the multiple non-uniform sub-regions and the intensity inhomogeneities. In this paper, we present a novel Region of Interest (ROI) detection method to minimize the ROI in the images automatically. We applied the geometric active contours that forces the variational level set function to be close to object boundaries. In addition, the mean-shift algorithm was used to reduce the sensitivity of parameter change in variational level set equation. In order to achieve the experiment, varieties of natural images in different modalities were tested. We compared the efficiency of the proposed method with the method using the human segmentation of the images. In a less complex background, the precision and recall are 92.77% and 88.95%, respectively. In a complex background, the precision and recall are 88.93% and 89.10%, respectively. The experimental results show that our method is imitating human decision making for ROI detection and evaluation. Key-Words: Image Segmentation, Active Contour, Region of Interest, Object of Interest, Level Set, Mean Shift Algorithm