Saliency Detection Framework Using Low Level Multi Scale Image Features Based on Wavelet Transformation Technique

Abstract

Stationary wavelet transform (SWT) achieves translation-invariance by avoiding down sampling, resulting in giving lossless information. In this paper, SWT technique is implemented for creating local and global saliency maps. The performance of the saliency detection method using low level features based on Discrete wavelet Transform (DWT) and SWT are analyzed for biomedical applications. The results are also extended for a hybridized saliency map. Finally the simulation results are analyzed and compared using different parameters for local, global and hybrid saliency detection models.