Salient region detection has been extensively studied due to its great possibilities for various computer vision fields. Despite remarkable research advances, a considerable amount of efforts still has been devoted to detect salient regions, which attract the human visual attention indeed, over the last few years. This is because previous methods are easily biased toward edges or corners, which are statistically significant, but not necessarily salient. Moreover, they often fail to find salient regions in complex scenes due to ambiguities between salient regions and highly textured backgrounds. In this thesis, we present a novel framework for salient region detection based on textural contrast, which is defined by combining the difference of luminance and directional coherence between center and surrounding regions. The proposed method is simple, robust, yet biologically plausible and it can thus be easily extended to a wide range of computer vision applications. Based on various data sets, we conduct comparative evaluations both qualitatively and quantitatively using 12 representative saliency detection models presented in literature, and the results show that the proposed scheme outperforms other previously developed methods in detecting salient regions. We further provide the utility of the proposed method by applying it to real-world applications such as content-aware image resizing (i.e., image retargeting), object segmentation, and video surveillance.