摘要：Over the last two decades, digital video technology has witnessed an explosive growth in both the diversity of techniques and its range of applications. As a consequence, we are also witnessing a dramatic expansion of the stocks of digital content. This in turn, is generating a critical need for systems and tools able to automatically structure and understand visual information. Visual Information Engineering embraces research and development related to image processing, machine vision and visual communications. Recent trends have been directed towards investigating important interactions between visual information retrieval, visualisation and advanced coding and streaming of images and video. The aim is to exploit techniques developed over the years to enable interaction and more advanced multimedia applications. The objective of this special issue is to collate and disseminate few relevant achievements of researchers working on visual information engineering. High quality contributions addressing related theoretical and practical aspects were received. After a thorough review process a total of six papers out of 14 submissions were selected. The first paper addresses the challenging problem of automatic recognition of highlights from videos for content-based indexing and retrieval applications. Ren et al. propose techniques to solve this problem using knowledge supported extraction of semantics, and employing compresseddomain processing for efficiency. In the next paper Badii et al., introduce a framework for process management support in the context of media production and more specifically 3D cinematographic immersive and interactive production. The authors present a holistic approach to software engineering applied to media production modelling to ensure design correctness, completeness and effectiveness. The focus of the designed application is on information and metadata management throughout the process in a similar fashion to that achieved in Decision Support Systems (DSS) to facilitate well-grounded business decisions. The paper sets out the aims and objectives and the methodology deployed. In the third paper, Li and Liu present a novel cross-layer approach to multiuser H.264/AVC video stream transmission over wireless networks. They consider the joint optimisation of three layers of the protocol stacks: the application (APP) layer, the media access control (MAC) layer and the physical (PHY) layer. An objective function for cross-layer optimisation is developed and the authors intend to maximise the perceptual quality of the videos transmitted through the wireless network. The fourth paper targets the difficult and important problem of coding distortion assessment in video streaming applications. Yang et al. propose an efficient method employing a linear model for assessing the coding distortion using the quantization scale. Their method further exploits properties of the human visual system by taking into account the spatial and temporal masking effects, whose complexities are estimated by a rate-distortion model formulating their relationship with the coding bit-rate. The next paper addresses a specific application of visual information retrieval – text readability in CCTV. In this paper, Newbold and Gillam discuss an approach to automate video annotations based on an expanded consideration of readability that covers both text and cognitive factors. The aim is to select ontological elements that support wider ranges of user queries through limited sets of annotations derived automatically through the analysis of expert annotations of prior content. Using readability studies to improve the text, the authors suggest that the accessibility of semantic content and the terminology extraction process can be improved. In the last paper, Janjusevic et al. propose new approaches for browsing initialisation, exploration and content access of rushes archives. Exploring and navigation through raw footage is assisted by organising the video material using a meaningful structure and by adopting appropriate visualisation solutions. Un-annotated content is organised in hierarchical previews, while browsing is enabled by novel methods of random exploration and random content access to preview nodes. This special issue has assembled a small sample of papers originating from well-known research groups. The contributing authors were instrumental in the completion of the special issue and the Guest Editors would like to thank all of them. The anonymous referees played a key role in the review and selection process ensuring the special issue includes only the submissions of high technical quality.
