Salient Object Detection in Videos Based on SPATIO-Temporal Saliency Maps and Colour Features

Salient object detection in videos is challenging because of the competing motion in the background, resulting from camera tracking an object of interest, or motion of objects in the foreground. The authors present a fast method to detect salient video objects using particle filters, which are guided by spatiotemporal saliency maps and color feature with the ability to quickly recover from false detections. The proposed method for generating spatial and motion saliency maps is based on comparing local features with dominant features present in the frame. A region is marked salient if there is a large difference between local and dominant features. For spatial saliency, hue and saturation features are used, while for motion saliency, optical flow vectors are used as features. Experimental results on standard datasets for video segmentation and for saliency detection show superior performance over state-of-the-art methods. INTRODUCTION Modern day life has overwhelming amount visual data and information available and created every minute. This growth in image data has led to new challenges of processing them fast and extracting correct information, so as to facilitate different tasks from image search to image compression and transmission over network. One specie problem of computer vision algorithms used for extracting information from images, is to and objects of interest in an image. Human visual system has an immense capability to extract important information from a scene. This ability enables humans to focus their limited perceptual and cognitive resources on the most pertinent subset of the available visual data, facilitating learning and survival in everyday life. This amazing ability is known as visual saliency (Itti et al. (1998)). Hence for a computer vision system, it is important to detect saliency so that the resources can be utilized properly to process important information. Applications range from object detection or Content Based Image Retrieval (CBIR), face or human reidentication and video tracking. Motivation Saliency is the ability or quality of a region in an image to standout (or be prominent) from the rest of the scene and grab our attention. Saliency can be either stimulus driven or task specific. The former one is known as bottom-up saliency while the later species top-down saliency and leads to visual search. Bottomup saliency can be interpreted as a filter which allows only important visual information to grab the attention for further processing[1],[2]. In our work, we concentrate on bottom-up salient object detection. Saliency is a particularly useful concept when considering bottom-up feature extraction, since one must and what is significant in an image from the scene data alone. In such circumstances, the role of context becomes extremely important. That is to say that saliency can be described as a relative measure of importance. Hence, the bottom-up saliency can be interpreted as its state or quality of standing out (relative to other stimuli) in a scene.