With modern socio‐economic development, the number of vehicles in metropolitan cities is growing rapidly. Therefore, obtaining real‐time traffic volume estimates has a very important significance in using the limited road space and traffic infrastructure. In this study, the authors present a video‐based traffic volume and direction estimation at road intersections. To discriminate the vehicles from the remaining foreground objects, vehicle recognition is performed by training a deep‐learning architecture from a pre‐trained model. This method, called transfer learning, primarily circumvents the requirement of huge labelled datasets and the time for training the network. The video sequence is first detected for moving foreground regions or patches. The trained model is subsequently used to classify the vehicles. The vehicles are tracked, and trajectory patterns are clustered using standard techniques. The number and direction of vehicles are noted, which are later compared with the manually observed values. All experiments were performed on real‐life surveillance sequences recorded at four different traffic intersections in the city of Kolkata.

While H.264 is a well-established standard for video surveillance, its high-profile implementation, in particular, has the unique capabilities that pack more visual detail into a given bitrate. Several algorithms exist that detect moving objects from the H.264 bitstream, most of which end up incorrectly classifying the nonstationary or dynamic components (e.g., jitter camera, waving trees, ripples, etc.) of the background as foreground. Moreover, the coarse quantization schemes adopted at constrained bitrates pose new challenges for direct identification of moving objects/targets from the bitstream. This paper focuses on dynamic background modeling for the H.264 video encoded at very low bitrates. To this end, enhanced features of the fundamental coding unit, i.e., the macroblock (MB) are proposed. Based on the temporal statistics of the MB features gathered from initial frames, MBs that potentially contain parts of moving objects are selected in subsequent frames. The selected MBs constitute a coarse segmentation map of the foreground at the MB level. Finally, pixel-level segmentation of the foreground is obtained by comparing pixels constituting the selected MBs with the colocated counterparts of a background frame. Experimental results showing the comparison of bitstreams encoded at strikingly low bitrates are obtained over a diverse set of standardized surveillance sequences.

While surveillance video is the biggest source of unstructured Big Data today, the emergence of high-efficiency video coding (HEVC) standard is poised to have a huge role in lowering the costs associated with transmission and storage. Among the benefits of HEVC over the legacy MPEG-4 Advanced Video Coding (AVC), is a staggering 40 percent or more bitrate reduction at the same visual quality. Given the bandwidth limitations, video data are compressed essentially by removing spatial and temporal correlations that exist in its uncompressed form. This causes compressed data, which are already de-correlated, to serve as a vital resource for machine learning with significantly fewer samples for training. In this paper, an efficient approach to foreground extraction/segmentation is proposed using novel spatio-temporal de-correlated block features extracted directly from the HEVC compressed video. Most related techniques, in contrast, work on uncompressed images claiming significant storage and computational resources not only for the decoding process prior to initialization but also for the feature selection/extraction and background modeling stage following it. The proposed approach has been qualitatively and quantitatively evaluated against several other state-of-the-art methods.

The H.264/Advanced Video Coding (AVC) is the industry standard in network surveillance offering the lowest bitrate for a given perceptual quality among any MPEG or proprietary codecs. This paper presents a novel approach for background subtraction in bitstreams encoded in the Baseline profile of H.264/AVC. Temporal statistics of the proposed feature vectors, describing macroblock units in each frame, are used to select potential candidates containing moving objects. From the candidate macroblocks, foreground pixels are determined by comparing the colors of corresponding pixels pair-wise with a background model. The basic contribution of the current work compared to the related approaches is that, it allows each macroblock to have a different quantization parameter, in view of the requirements in variable as well as constant bit-rate applications. Additionally, a low-complexity technique for color comparison is proposed which enables us to obtain pixel-resolution segmentation at a negligible computational cost as compared to those of classical pixel-based approaches. Results showing striking comparison against those of proven state-of-the-art pixel domain algorithms are presented over a diverse set of standardized surveillance sequences.