Locally Adaptive Passive Error Concealment for Wavelet Coded Images

Error Concealment for 3-D DWT Based Video Codec Using Iterative Thresholding

“…At the decoder side, we recover the video sequence utilizing a basic reconstruction where all lost coefficients are estimated as zeros (Fig. 1 a)), using linear interpolation [9] (Fig. 1 b)) and using IST with σ 0 = 200, K = 40 and H × W discrete cosine transform (DCT) as the sparsifying transform ( Fig.…”

“…This approach gives relatively good reconstruction performance, especially if the lowest subbands coefficients are delivered using additional protection, such as duplication of lowest subband coefficients [3]- [5], [7], multiple description coding [8] and/or inter-packet forward error-correction [3]- [5]. If neighbor coefficients are delivered, then a lost coefficient can be interpolated [6], [9] utilizing local correlation of a subband coefficient. This recovery method can improve the reconstruction quality without significant increase of computational complexity, but it cannot be efficiently used when many or even all coefficients of a subband are lost.…”

Passive Error Concealment for Wavelet-Coded I-Frames With an Inhomogeneous Gauss–Markov Random Field Model

“…Compared to more common block-based approaches, relatively few passive concealment methods were reported for wavelet coded images and video [2], [5]- [7]. For the reconstruction of lost low-frequency coefficients (the scaling coefficients), the existing methods are mainly traditional error concealment algorithms from the image domain which are slightly adapted to be able to work in the wavelet domain.…”

“…This technique gives slightly better results than bilinear interpolation, but it requires much more computational effort. In our previous work [7], the lost low-frequency coefficients are reconstructed with a locally adaptive interpolation scheme. For each lost coefficients, the interpolation weights are estimated based on the interpolation errors that arise from interpolating nearest correctly received neighbors in both horizontal and vertical directions.…”

“…For each lost coefficients, the interpolation weights are estimated based on the interpolation errors that arise from interpolating nearest correctly received neighbors in both horizontal and vertical directions. While this method of [7] interpolates the lost data from the immediate horizontal and vertical neighbors like the proposed method, the concept behind determining the weights is totally different. In our earlier method [7], we derive a closed form expression under the side constraint that the sum of the horizontal and vertical interpolation weights equals 1.…”