Towards visualization and searching :a dual-purpose video coding approach

Abstract

In modern video applications, the role of the decoded video is much more than filling a screen for visualization. To offer powerful video-enabled applications, it is increasingly critical not only to visualize the decoded video but also to provide efficient searching capabilities for similar content. Video surveillance and personal communication applications are critical examples of these dual visualization and searching requirements. However, current video coding solutions are strongly biased towards the visualization needs. In this context, the goal of this work is to propose a dual-purpose video coding solution targeting both visualization and searching needs by adopting a hybrid coding framework where the usual pixel-based coding approach is combined with a novel feature-based coding approach. In this novel dual-purpose video coding solution, some frames are coded using a set of keypoint matches, which not only allow decoding for visualization, but also provide the decoder valuable feature-related information, extracted at the encoder from the original frames, instrumental for efficient searching. The proposed solution is based on a flexible joint Lagrangian optimization framework where pixel-based and feature-based processing are combined to find the most appropriate trade-off between the visualization and searching performances. Extensive experimental results for the assessment of the proposed dual-purpose video coding solution under meaningful test conditions are presented. The results show the flexibility of the proposed coding solution to achieve different optimization trade-offs, notably competitive performance regarding the state-of-the-art HEVC standard both in terms of visualization and searching performance.