Table 2 Enhancements to TDWZ
Module | Summary | References |
|---|---|---|
Rate-control | Rate-allocation scheme without using feedback channel and without significantly increasing the encoder complexity A linear rate estimation model that avoids over or under estimation and achieves optimal rate-distortion performance Shows a reasonably good encoder rate allocation performance while maintaining coding efficiency | (Sheng et al. 2010) |
Encoder based rate-allocation scheme that predicts the number of encoding bits as a function of the coding mode and quantization parameters Predictions select the best coding mode and quantization parameters for encoding WZ frame without significantly increasing encoder’s complexity Shows relatively low loss in rate-distortion performance as compared to conventional decoder based rate allocation scheme | (Sheng et al. 2008) | |
The DISCOVER codec generates the band of transform coefficients after performing block-based transformation and quantization on each WZ frame An improved rate-control algorithm that computes the initial number of bits to transmit for each bitplane and band Does not require excessive number of iterations, hence improves the coding efficiency | (Artigas et al. 2007) | |
A feedback channel-driven rate-control codec with improved modules using special coding tools An improved motion learning algorithm with reasonably good rate-distortion performance | (Martins et al. 2010) | |
A hybrid procedure for rate estimation at encoder side for wavelet based WZ codec with no feedback channel Quantized high resolution subbands entropy coded using low-complexity intra-coding to avoid under-estimation of required bit rate For lower resolution subbands, the entropy of the bitplane crossover probability is used as an estimate | (Bernardini et al. 2011) | |
Channel adaptive rate control for feedback-free wavelet-based WZ; priority given to lower frequency subbands that hold more information about a frame than higher frequency subbands | (Rui et al. 2013) | |
Decoding algorithm | A decoding algorithm that exploits previously reconstructed transform bands to reduce the total number of bits needed for reconstructing the remaining bands | (Martins et al. 2010) |
Presents TRACE (TRansform domain Adaptive Correlation Estimation) for WZ decoder Progressively learns the correlation among frames during the frame reconstruction process A convex optimization based band-level correlation estimation method that minimizes the theoretical required bit rate | (Fan et al. 2010) | |
A multi-hypothesis based WZ decoder that exploits the redundancy between multiple side information(s) and the source frame Uses both block-based and optical flow-based side information generation methods to generate multiple side information(s) An optical flow based frame interpolation algorithm to compensate side information estimation weaknesses in block based methods Employs multiple soft-inputs for decoding and reconstruction based on a weighted joint distribution, which reduces the required bitrate and improves the quality of reconstructed frames | (Huang et al. 2011) | |
A bit-level context-adaptive correlation model for the decoder of a wavelet-based WZ codec with MRMR (Liu et al. 2010) to achieve better prediction of the bit probability distribution. | (Qing and Zeng 2014) | |
Side information | An improved side information refinement framework that utilizes both spatio-temporal correlation and previously reconstructed transform bands Uses a non-local means denoising process that exploits partially decoded side information(s) progressively available at the decoder A denoising process that progressively regenerates side information with improved quality | (Shen et al. 2012) |
A side information generation algorithm based on semi super-resolution frame that exploits past and future reference frames for block-based motion estimation process Iteratively generates side information using channel decoder following the decoding of a low resolution base layer to produce high quality decoded frame Low resolution encoding reduces the encoding complexity Encodes the residual WZ frame by cosets | (Macchiavello et al. 2007) | |
An extrinsic information transfer (EXIT) chart analysis for minimizing the mutual information variation in iterative low density parity check (LDPC) decoding during side information refinement process Obtains relatively good reconstruction quality with low coding rates | (Wen et al. 2012) | |
Correlation noise modeling | A correlation noise estimation algorithm that exploits the adjacent key frames to predict the quantization step of the quantization module Indices generated for two frames differ by smallest number of bits possible Shows improved WZ encoding bitrate for a given target quality as compared to the traditional coding schemes | (Micallef et al. 2012) |
An online approach for modelling correlation noise model parameters at the decoder Determines the temporal correlation between frames at various levels of granularity (DCT bands, DCT coefficients) Shows good rate distortion performance at higher estimation granularity | (Brites and Pereira 2008) | |
A correlation noise model that utilizes cross-band correlation to precisely estimate the Laplacian parameters Uses a category map based on previously reconstructed bands to classify transformed coefficients of the current band Allocates each transformed coefficient to a Laplacian parameter based on its category More precise estimation of correlation noise model parameters leads to better rate-distortion performance | (Huang and Forchhammer 2009) |