PROJECT TITLE :

Salient Region Detection via Integrating Diffusion-Based Compactness and Local Contrast - 2015

ABSTRACT:

Salient region detection could be a challenging downside and an vital topic in laptop vision. It incorporates a big selection of applications, such as object recognition and segmentation. Several approaches have been proposed to detect salient regions using different visual cues, like compactness, uniqueness, and objectness. However, each visual cue-based mostly methodology has its own limitations. After analyzing the benefits and limitations of various visual cues, we have a tendency to found that compactness and native contrast are complementary to each other. Also, native contrast can very effectively recover incorrectly suppressed salient regions using compactness cues. Motivated by this, we tend to propose a bottom-up salient region detection technique that integrates compactness and native contrast cues. Furthermore, to produce a pixel-correct saliency map that a lot of uniformly covers the salient objects, we tend to propagate the saliency information using a diffusion process. Our experimental results on four benchmark information sets demonstrate the effectiveness of the proposed methodology. Our methodology produces additional correct saliency maps with higher precision-recall curve and better F-Live than different 19 state-of-the-arts approaches on ASD, CSSD, and ECSSD data sets.