Nonlocal center-surround reconstruction-based bottom-up saliency estimation

Abstract

The center-surround comparison principle is widely used in existing bottom-up saliency estimation models. However, most of them are based on local image processing techniques which are hard to handle texture regions well as a relatively large neighborhood is required to represent textures. In this paper, we propose a nonlocal patch-based reconstruction approach to reformulate the center-surround comparison. In the proposed approach, the saliency is measured by the reconstruction residual of representing the central patch with a linear combination of its surrounding patches. As a generalization of Itti et al.'s classical center-surround comparison scheme, the proposed approach performs well on images with symmetric structures where Itti et al.'s method fails, as well as on general natural images. Numerical experiments show the proposed approach produces better results compared to the state-of-the-art algorithms on several public databases.

A novel approach to reconstruction based saliency detection via convolutional neural network stacked with auto-encoder

“…Thirdly, compared with other reconstruction strategies which use linear combination or auto-encoder network [6,7], the CNNR model reconstructs the saliency Our proposed deep reconstruction model, CNNR, has the following strengths:…”

“…Reconstruction strategies have been presented to predict saliency region. In Xia et al [6] the reconstruction of the patch was realized by a linear combination, and salient regions were obtained by estimating the difference between the reconstructed patch and the original patch. In Ren et al [28] a regularized feature reconstruction framework was presented to highlight salient regions for video.…”

“…Reconstruction based saliency detection is to emulate this ability. In [6], Xia et al proposed a reconstruction model to detect salient regions by computing the sparse reconstruction residual of central patches. The reconstruction process is realized by a linear combination of surrounding patches.…”