An Unsupervised Salient Object Extraction Approach using Statistical Method and High Contrast Saliency Map
Tallha Akram, Qichang Duan, Hongying Xu, Atif Amin · 2013
This paper proposes a multilevel unsupervised image segmentation approach aimed at salient object extraction. The fundamental objective is the extrication of pronounced object with confined boundary. Taking full advantage of HSV color space, evident of being nearest colorspace to human perception, proposed methodology segment the image using Expectation Maximization (EM), to estimate the parameters of Gaussian Mixture Model (GMM), using heuristic initialization. The Binary Partitional Tree (BPT) then extract the prime focus from reduced color palette on the basis of largest eigenvector. The final amalgamation with high contrast saliency Map diminishes all inutile fragments and excerpt salient object with ensured boundary. The experimental data indicates that hybrid approach leads to improved color segmentation with the apparent assertion of prime object extraction.