A Video Saliency Detection Model in Compressed Domain

Saliency detection in MPEG and HEVC video using intra-frame and inter-frame distances

“…The main differences between the proposed work and that reported in [19] is as follows. In this work, we apply the concept of center-surround differences to compute feature vectors based on spatial, temporal and global differences.…”

“…This work uses the concept of center-surround differences in the computation of saliency maps [19]. The basic idea is to compute the distances between a feature vector representing a MB/CU and its surrounding feature vectors.…”

“…Location of targets of interest can also be detected [18]. Lastly, the authors in [19] proposed a video saliency detection model based on the concept of center-surround feature differences in the compressed domain. Block-based features are extracted from motion information, luminance, color and texture.…”

Detecting and Removing Visual Distractors for Video Aesthetic Enhancement

“…Dong et al [14] proposed a method to find salient objects in the wide luminous range of HDR videos. To facilitate the saliency analysis in compressed videos, Fang et al [37] utilized the feature sets extracted from the compressed domain to detect the salient regions. In general, video saliency analysis provides useful cues to the task of distractor detection.…”

“…Previous video saliency methods [33], [37] can generate framebased saliency maps that are temporally smooth. However, for our task, besides rough distribution maps of saliency, we want to know how distinctive each TSP element is.…”