More and more people are getting access to 3D; 3D-TV is a next expecting step in telecommunication. Before being presented to the public, any 3D content has to be coded, compressed and transmitted. All these treatments can impact the quality of the final product, thus it is essential to have a measurement tool for the estimation of the quality of stereoscopic content. Several studies have already modified the existing 2D quality metrics to be able evaluate 3D images, however the results are not satisfying. In this work, we propose a full-reference metric for a quality assessment of stereoscopic images employing the properties of binocular perception. The principle of the metric estimates the probability of fusion, that can be obtained perceptually. The quality of one view is assessed relatively to the other, and according to the result metric makes a decision, that is based on the binocular fusion properties. The comparison between views is performed only on the salient area. It is detected using the visual attention model based on the monocular depth cues and interest points. The metric has been tested on the publicly available dataset, and its results are coherent to the subjective scores.