A Robust Modeling of Impainting System to Eliminate Textual Attributes While Restoring the Visual Texture from Image and Video

摘要

The area of video analytics in the context of collaborative networking has gained a lot of attention from the research community owing to its potential applicability in the real life aspects. However, although image and video content which mostly get exchanged in the networking pipelines consist of several significant textual information from the application view-point which often display various confidential textual credentials of a corresponding individual. The realization of this fact that this textual attributes has to be removed for various image forensic requirements, has led to image impainting. The study has addressed this problem and come up with a novel analytical solution which imposes two different methods and further combines this two. In the 1~(st) stage it applies a robust mechanism to detect the region of an image and video frame sequence where textual data representation can be localized and perform extraction of those data it introduces artifact and visual anomalies. On the completion of this stage in the 2~(nd) phase, to eliminate the artifacts from the respective locations, it introduces a novel impainting technique which is computationally efficient and attain higher degree of textual data eliminated recovered image or video sequence which is almost similar like the original image or video sequence, can be visually perceived. The comparative performance analysis show that the proposed technique attain better outcome in terms of textual attributes detection accuracy (%) from specific region of interest (ROI) and also consume very less processing time (Sec) in contrast with the existing system.