Minimizing a weighted error criterion for spatial error concealment of missing image data

Frequency selective signal extrapolation with applications to error concealment in image communication

“…In case of concealment, the isotropic weighting function gains severable dBs in PSNR [14] with respect to a binary weighting function used in [13]. The concealment performance decreases if there are details in the support area which do not belong to the missing area.…”

“…As already mentioned in Section 2, the weighting function (2) should emphasize pixels which are more important for the extrapolation over less important ones. Therefore, we choose the isotropic model [14] for [m, n]…”

“…The parameters are chosen based on the investigations done in [14], where also PSNR results with respect to the concealment techniques [8][9][10][11][12] are published. The support area is 13 pixels in each direction.…”

Spatial error concealment with sequence-aligned texture modeling and adaptive directional recovery

“…To improve the performance of [26], a couple of algorithms were proposed to minimize variations along edges or geometric structures of corrupted image regions. This is often achieved by detecting lost image textures in the image recovery process [2,6,[8][9][10][11][12]16,17,19,23,24,27,32,33,35]. Some of them are briefly reviewed below.…”

“…Li and Orchard [11] used the orientation-adaptive interpolation and the linear merge strategy to recover lost pixels. Meisinger and Kaup [12] used the frequency extrapolation technique to perform the spatial error concealment. Kumwilaisak and Hartung [10] employed the technique in bioinformatics to estimate texture globally and then interpolated lost pixels geometrically in their error concealment algorithm.…”