Time-frequency-tuned salient region detection and segmentation

The present paper describes an efficient method for detecting and segmenting salient region(s) in an image. The method uses a time-frequency tuned salient region extraction technique based on wavelet transform (WT). WT provides both spatial and spectral characteristics (i.e., texture information) of pixels and hence can be utilized effectively for improving quality of salient region detection. As a result, the proposed method generates full resolution maps with uniformly highlighted regions with well defined boundaries, and invariant to translation, rotation and scaling that make it more useful in applications like object segmentation / recognition and adaptive compression. The superiority of the proposed method over the existing, is demonstrated both qualitatively and quantitatively using the indexes like precision and recall with a large set of benchmark data sets.