### Table 2: Results of hyperparameter estimation and dis- tance between the original and estimated images for the two methos JMAP and EM-MMAP.

"... In PAGE 3: ...eously estimated the hyperparameters ( 1; 2). Fig. 3 and Fig. 4 show the restored images by JMAP and EM-MMAP methods and the Table2 . summarizes the... ..."

### Table 1. Values of auto cumulants of images used as inde- pendent components. For each image, true values calculated from original images are in the upper row, and estimates are in the bottom row. a6 a16 :a building, a6 a17 :fabric, a6 a5 :tiles.

"... In PAGE 4: ...2. The cumulants of original image data and their estimates by a34 a44 a0 a48 a38 a7 appear in Table1 , and the estimation is good. Histograms of original image data appear in Figure 2.... ..."

### Table 1. Summary of the Quality Index Q of the filtered image after using different filters

2002

"... In PAGE 8: ... This image yields an excellent estimate of figure 1. The quality indexes achieved by the different filters are compared in Table1 . Figure 8 illustrates the trend of quality indexes Q, Q1, Q2, and Q3 versus iteration number.... ..."

### Table 1: Results of error diffusion kernel estimation using the original gray scale images and their error diffused version. The true kernel used was the Floyd-Steinberg kernel.

"... In PAGE 20: ... In other words, we use an original gray scale image and its error diffused version as the inputs for estimating the error diffusion kernel. Table1 shows the result of the estimation for both the pepper and Lena images where the Floyd-Steinberg kernel was the true kernel used. It can be observed from Table 1 that convergence was essentially achieved after two passes through the image.... In PAGE 20: ... Table 1 shows the result of the estimation for both the pepper and Lena images where the Floyd-Steinberg kernel was the true kernel used. It can be observed from Table1 that convergence was essentially achieved after two passes through the image. Furthermore, the differences between... In PAGE 22: ... Despite the fact that we only used an estimated gray scale image in the kernel estimation process, Table 2 shows that accuracy of the estimated coefficients are still quite accurate. The fact that it takes more passes (processing) for convergence when compared to the results in Table1 is perhaps not unexpected because we only use an estimated gray scale image in this case. Experimental results with many test images so far indicate that convergence can always be achieved.... ..."

### Table 1 Original and estimated surface

"... In PAGE 8: ... derived granulometric estimators using the thirteen previously described by partitioning each image parts and using three sections of each for training, total of 21 training images. The designed estimators been applied to the left-out lower right quarter age for comparison with the classical roughness The results are shown in Table1 . For samples and 6, the results are good, with outstanding samples 1 and 6.... ..."

### Table 1: Error in estimate relative to the original head phantom.

2000

"... In PAGE 11: ...Table 1: Error in estimate relative to the original head phantom. Table1 shows error measures relative to the original image, to provide context for the magnitude of differences between the MAP and ICD/GN estimates of the head phantom appearing in Table 2. The mag- nitudes of the differences displayed in Table 2 for different numbers of expansion updates show that the results are sufficiently accurate after one or two updates of the expansion first computed from the FBP im- age for ICD/GN to be very similar to an exact MAP reconstruction.... ..."

### Table 5.2: Mean square error table. A smaller mean square error indicates that the estimate is closer to the original image. The numbers have to be compared on each row. The square of the number on the left-hand column gives the real variance of the noise. By comparing this square to the values on the same row, it is quickly checked that all studied algorithms indeed perform some denoising. This is a sanity check! In general, the comparison performance corroborates the previously mentioned quality criteria.

### Table 5 Translation parameter and scaling factor estimated by Stirmark Angle X translation Y translation Estimate angle X estimate translation Y estimate translation

"... In PAGE 9: ...5 Experimental results with translation parameter and scaling factor estimation No matter what order of geometric distortions of translation and scaling is done to the watermarked image, we estimate the translation parameter and scaling factor with Tchebichef moments of the original image. Table5 gives the estimation results of translation and scaling estimation results. The attacks are given by Stirmark and the estimation results have a high precision.... ..."

### Table 1: Estimated facial animation parameters in comparison to the correct values. reconstruct the sequence by rendering the 3D mo- del. First, only the global motion (head transla- tion and rotation) is estimated and then all FAPs are determined. In Figure 4 the PSNR between the resulting reconstructed image and the original one is plotted over time. The high values of about 70 dB show that the images are nearly identical.

1997

"... In PAGE 4: ... The viewing angle of the camera is about 25o. The result of one such experiment can be seen in Table1 . Ten anima- tion parameters are changed, the others remain constant.... ..."

Cited by 8

### Table 3. Summary of timing estimates for derivative estimator

in By

2001

"... In PAGE 10: ...able 2. Data transfer timing for WildCard/E IO modes ................................................36 Table3 .... In PAGE 50: ... 38 would then be the sum of the data transfer time (at DMA rates) plus the element processing required for the five derivatives. Table3 summarizes these performance estimates by design, identifying the original subroutine ( Software ), the performance of basic core which does meet the FM timing model using register-mode ( Core reg ) and programmed-IO mode ( Core PIO ) data transfers, a design which computes both first and second derivative terms at once ( 2-deriv ), and the optimum hardware design discussed above ( Optimum ). Table 3.... ..."