Detection of salient object using pixel blurriness
Yi‐Chong Zeng, Chi‐Hung Tsai · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference · 2012
In this paper we propose a method to detect salient object in still image and non-slow motion background video. The key technique is measuring pixel blurriness. Generally speaking, salient object was taken in focus, pixels within salient object should be sharper than those within background. In the first step image intensity is extracted, and then four different-size average filters are applied to intensity. Subsequently, variation of intensity differences (VID) is computed among the original intensity and four blurred versions. The VID is employed to represent degree of pixel blurriness. Finally, a thresholding method is applied to pixels' blurriness in order to distinguish salient object from background, and salient object is composed of low-blurriness pixels. The experiment results demonstrate that the proposed method is efficient in detection of salient object in still image and non-slow motion background video. Moreover, our method has better detection performance than the two compared methods.