Detection of salient objects in image and video is of great importance in many computer vision applications. In spite of the fact that the state of the art in saliency detection for still images has been changed substantially over the last few years, there have been few improvements in video saliency detection. This paper investigates the use of recently introduced non-local neural networks in video salient object detection. Non-local neural networks are applied to capture global dependencies and hence determine the salient objects. The effect of non-local operations is studied separately on static and dynamic saliency detection in order to exploit both appearance and motion features. A novel deep non-local neural network architecture is introduced for video salient object detection and tested on two well-known datasets DAVIS and FBMS. The experimental results show that the proposed algorithm outperforms state-of-the-art video saliency detection methods.

Detecting car motion in video frames is one of the key subjects in computer vision society. In recent years, different approaches have been proposed to address this issue. One of the main challenges of developed image processing systems for car detection is their shadows. Car shadows change the appearance of them in a way that they might seem stitched to other neighboring cars. This study aims to propose an optimized method for removing car shadows using entropy and Euclidean distance features. For each pixel, a weight is assigned according to the mentioned features. The weights assigned to shadows and background (asphalt) pixels are very close to each other which enable the background subtraction to remove both of them. The proposed method was evaluated on three datasets based on OA, HR, FAR, MODP and MOTP measures. The method was also compared with both NCC and HSV color methods which are well-known in removing car shadows. The results showed that the proposed methods depending on the type of the index is variable between 3 to 12 percent accurate results.