In this paper, a novel method for salient object detection from natural images is proposed. In order to extract the object from an image which is visually attractive, non-redundancy is conceptually incorporated to define the saliency of image pixels by applying principal component analysis (PCA) to color components of an image. From the principal component images, seed pixels for object and background are extracted. Using these object and background seed pixels as training samples, linear discriminant analysis (LDA) is applied to image pixels so that the pixels are classified as object or background. Experiments on test images show that not only the performance of the proposed method is promising, but also it works competitively with state-of-the-art salient object detection methods.
CITATION STYLE
Lee, H., Kim, J., & Kim, J. (2014). Automatic salient object detection using principal component analysis. In Advances in Intelligent Systems and Computing (Vol. 274, pp. 717–723). Springer Verlag. https://doi.org/10.1007/978-3-319-05582-4_62
Mendeley helps you to discover research relevant for your work.