Spatial Shape Error Concealment Utilising Image Texture

In applications where shape is used as metadata to describe image and video content, image dependent techniques exhibit superior error concealment performance compared to the traditional MPEG-4 like image-independent techniques. The image corresponding to a particular shape may not be readily known at the decoder and rather needs to be retrieved from of a set of images. This paper presents a novel shape error concealment technique utilising image texture (SECI) that investigates image matching techniques based on partial-shapes and use the obtained image for error concealment. Iterative closest point, independent component analysis and the modified Hausdorff distance algorithms are used to establish the corresponding image to the shape. The rubberband contour detecting function is employed in the core of the error concealment process. Experimental results show the effectiveness of the proposed technique and corroborate the superior performance of the new SECI technique compared with established shape-based concealment techniques.