Image Dependent Spatial Shape Error Concealment for Multiple Shapes

Existing shape error concealment techniques consider a single closed shape at a time. However, in many applications there are commonly multiple shapes in a scene/object and they are not necessarily closed. Existing techniques attempt to conceal errors whenever they find a decoded shape is broken. However, in a multiple shape context, it is not that straightforward, as it becomes crucial to determine whether a segment is broken due to data losses or just the beginning of a new segment. This paper presents an image dependent shape-error concealment technique for multiple shapes (ISCM) which exploits textural information using a rubberband function to determine proper localisation of the shape errors and effectively recover them. Comparative experimental results analysis confirms both a superior error concealment performance and an improved robustness of ISCM technique.