Fast and Efficient Saliency Detection Using Sparse Sampling and Kernel Density Estimation

Abstract

Abstract. Salient region detection has gained a great deal of attention in computer vision. It is useful for applications such as adaptive video/image compression, image segmentation, anomaly detection, image retrieval, etc. In this paper, we study saliency detection using a center-surround approach. The proposed method is based on estimating saliency of local feature contrast in a Bayesian framework. The distributions needed are estimated particularly using sparse sampling and kernel density estimation. Furthermore, the nature of method implicitly considers what refereed to as center bias in literature. Proposed method was evaluated on a publicly available data set which contains human eye fixation as groundtruth. The results indicate more than 5% improvement over state-of-theart methods. Moreover, the method is fast enough to run in real-time.

Do Saliency Models Detect Odd-One-Out Targets? New Datasets and Evaluations

“…Saliency Models and Datasets. For all experiments that follow we use 12 classical saliency models (AIM [12], BMS [73], CVS [20], DVA [28], FES [51], GBVS [26], IKN [32], IMSIG [29], LDS [21], RARE2012 [43], SSR [46] and SUN [74]) and 8 deep learning models (DGII [39], DVAP [59], eDN [58], ICF [40], MLNet [18], oSALICON [30,52], SalGAN [41] and SAM-ResNet [19]). We use the SMILER framework [64] to run all models without center bias on P 3 and with center bias on O 3 .…”

An improved saliency model of visual attention dependent on image content

“…We use the metric of AUC-Borji and sAUC to compare our model to the models shown in Figure 4, with AUC scores ranging from low to highest reported scores. We use the IttiKoch model (Walther and Koch, 2006) as an extension of the base model of Itti et al (1998), the Achanta model (Achanta et al, 2009) as one of the most cited models in the frequency domain, and the SUN saliency model (Zhang et al, 2008) and the Fast and Efficient Saliency (FES) model (Tavakoli et al, 2011) as two of the Bayesian-based models. We use the Murray model (Murray et al, 2011) that uses wavelet transform to generate scales like our model.…”

A DCA‐based sparse coding for video summarization with MCP

“…To integrate the center bias into the saliency mapping framework, we utilize the object‐biased Gaussian refinement process [18, 41, 42] to generate the center prior. To each pixel z whose coordinate is $$(x, y)$$, we denote the salient mapping of the center prior as: $$$$\begin{equation} \mathit {Sal}^+(z) = \exp {\left[-{\left(\frac{(x_z-x_\mathrm{obj})^2}{2\sigma _\mathrm{x}^2}+\frac{(y_z-y_\mathrm{obj})^2}{2\sigma _\mathrm{y}^2}\right)}\right]}, \end{equation}$$$$where σ x , σ y are set as $$\frac{1}{4}$$ the image's height and width and x obj , y obj are the object centers, defined as: $$$$\begin{align} x_\mathrm{obj} = \sum _{z=1}^N \omega _z x_z,\quad y_\mathrm{obj} = \sum _{z=1}^N \omega _z y_z, \end{align}$$$$where, $$\omega _z$$ is the weight, which are defined by the reconstruction error on pixel level $$E(z)$$ in (): …”