### Table 4. Fused Gaussians example

"... In PAGE 16: ... For this example, we start with a uniform sample over [727 7]2 and the target density on iteration t is 13t4x7 y5 29 16x16 f04x7 y5. In Table4 , we give the minimum, maximum and aver- age estimated relative errors over ten trials, as well as the smallest, largest and average point estimates for each of the three experimental setups. We find that the trial that gave the smallest estimated relative error was (1)1 Figure 6 depicts the contour plot.... ..."

### Table 2: Comparison of scheme of [8] and [9] for Gaussian source.

"... In PAGE 11: ... The curve in Fig.3 (extracted from [14]) are based on using a trellis with 32 states for the packing, resulting in an increase by a factor of about 32 for the complexity of the TB-SVQ as compared to the values given for [8] in Table2 . Referring to Tables 2, 3, and including the increase in the complexity due to the packing trellis for TB-SVQ [14], we can obtain an estimate of the reduction in complexity for the proposed method as compared to [14] while the performances are very close (refer to Fig.... ..."

### Table 3: Estimated 5 5 2D gaussian PSF, 40 iterations, red channel, decoupled EM

1995

"... In PAGE 33: ... #1: degraded RGB Lena, intra-channel blurs only, 20 dB SNR Figure 4: Exp. #1: Restored RGB (top to bottom) Lena: decoupled EM, multi-channel EM, and iterative multi-channel Wiener (left to right) Table 1: 2D gaussian PSF, 5 5 support Table 2: SNR Improvement (dB), no cross channel blur Table3 : Estimated 5 5 2D gaussian PSF, 40 iterations, red channel, decoupled EM Table 4: Estimated 5 5 2D gaussian PSF, 40 iterations, red channel, multi-channel EM Table 5: SNR improvement (dB) for experiment # 2 Figure 5: Exp. #2: Degraded red, green, and blue images, cross-channel degradations Figure 6: Exp.... ..."

### Table 2: Comparison of scheme of [9] and [10] for Gaussian source.

"... In PAGE 11: ...using a trellis diagram for packing results in a linear increase in the complexity proportional to the number of states for the method of [9]. The curve in Figure 3 (extracted from [15]) are based on using a trellis with 32 states for the packing, resulting in an increase by a factor of about 32 for the complexity of the TB-SVQ as compared to the values given for [9] in Table2 . Referring to Tables 2 and 3, and including the increase in the complexity due to the packing trellis for TB-SVQ [15], we can obtain an estimate of the reduction in complexity for the proposed method as compared to [15] while the performances are very close (refer to Figure 3).... ..."

### Table 2: Comparison of scheme of [9] and [10] for Gaussian source.

"... In PAGE 11: ...using a trellis diagram for packing results in a linear increase in the complexity proportional to the number of states for the method of [9]. The curve in Figure 3 (extracted from [15]) are based on using a trellis with 32 states for the packing, resulting in an increase byafactor of about 32 for the complexity of the TB-SVQ as compared to the values given for [9] in Table2 . Referring to Tables 2 and 3, and including the increase in the complexity due to the packing trellis for TB-SVQ [15], we can obtain an estimate of the reduction in complexity for the proposed method as compared to [15] while the performances are very close (refer to Figure 3).... ..."

### Table 4: Estimated 5 5 2D gaussian PSF, 40 iterations, red channel, multi-channel EM

1995

"... In PAGE 33: ... #1: degraded RGB Lena, intra-channel blurs only, 20 dB SNR Figure 4: Exp. #1: Restored RGB (top to bottom) Lena: decoupled EM, multi-channel EM, and iterative multi-channel Wiener (left to right) Table 1: 2D gaussian PSF, 5 5 support Table 2: SNR Improvement (dB), no cross channel blur Table 3: Estimated 5 5 2D gaussian PSF, 40 iterations, red channel, decoupled EM Table4 : Estimated 5 5 2D gaussian PSF, 40 iterations, red channel, multi-channel EM Table 5: SNR improvement (dB) for experiment # 2 Figure 5: Exp. #2: Degraded red, green, and blue images, cross-channel degradations Figure 6: Exp.... ..."

### Table 1: Gaussian errors

"... In PAGE 25: ...25 The Monte Carlo results lead to four preliminary conclusions: i) The ARCH parameters (! and ) are very badly estimated by OLS. This in- e ciency is more and more striking when one goes from Table1 to Table 3. While the het- eroskedasticity parameter is underestimated by OLS by almost 20 percent in the gaussian case, it is underestimated by almost 50 percent in the gamma case, that is when both leptokurtosis and skewness are present.... ..."

### Table 1: 2D gaussian PSF, 5 5 support

1995

"... In PAGE 16: ...We will see that the extra information contained in the cross-channel power spectra can add signi cant improvement to the restored image as well as a more accurate estimation of the blur. The rst experiment had an SNR = 20 dB and a 5 5 truncated Gaussian blur for each of the images, where the values of the PSF are given in Table1 . The original Lena image was used for this experiment, shown in Fig.... In PAGE 17: ...The values of the estimated PSF for the red channel for both the decoupled EM and multi-channel EM experiments are listed in Tables 3 and 4, respectively. In both cases, the estimated PSFs were very close to the true PSF, listed in Table1 . In the next subsection, we will show a much greater di erence in the estimates of the PSF between the decoupled and multi-channel EM approaches when cross-channel blurs are present.... In PAGE 33: ... #1: degraded RGB Lena, intra-channel blurs only, 20 dB SNR Figure 4: Exp. #1: Restored RGB (top to bottom) Lena: decoupled EM, multi-channel EM, and iterative multi-channel Wiener (left to right) Table1 : 2D gaussian PSF, 5 5 support Table 2: SNR Improvement (dB), no cross channel blur Table 3: Estimated 5 5 2D gaussian PSF, 40 iterations, red channel, decoupled EM Table 4: Estimated 5 5 2D gaussian PSF, 40 iterations, red channel, multi-channel EM Table 5: SNR improvement (dB) for experiment # 2 Figure 5: Exp. #2: Degraded red, green, and blue images, cross-channel degradations Figure 6: Exp.... ..."

### Table 5: Results for Gaussian variation sources.

2007

"... In PAGE 6: ... We also compare n2SSTA with our implementation of [2] (denoted as linSSTA) by assuming Gaussian variations and linear delay model for both. From Table5 , we see that in predicting = , n2SSTA matches Monte Carlo simulation well with about 5.5% error, while linSSTA has about 11% error.... In PAGE 6: ... This clearly shows that n2SSTA is not only more general, but also more accurate than linSSTA. Note that n2SSTA has a larger error for Gaussian variation sources in Table5 than for uniform or triangle variation sources in Table 4, and this is because n2SSTA needs bigger bounds (10) for Gaussian variations than for uniform or triangle variations. Interestingly, we nd that both approaches pre- dict the 95% yield point well.... ..."

Cited by 1

### Table 2 Computation results of the inhomogeneity indicator.

"... In PAGE 6: ...umination. The pressed cork texture was taken from Ref. 19, and a Gaussian inhomogeneity i~x! was superposed to it according to g~x!5g~x,y!5F2 x ~N21!DxG@t~x!1i~x!# 1const t~x!i~x! x ~N21!Dx . ~26! The computation results of the inhomogeneity indicator are shown in Table2 . Because of the lower harmonic dis- tortion, the computation of H2$%was performed in the frequency domain.... ..."