### Table 2 Examples of estimated and measured bit error rates for embedded watermarks. (y: no bit errors observed for a watermark of length 10000 bit.) =

"... In PAGE 10: ...) = 1 p2 1 r 2 1erfc 2 p cr mean( i) p2 1 ! (11) and nally BER = 12 erfc 0 @ p pcr mean( i) p2 q 2 v + 2 v 1 A : (12) At the same time, the data rate RWM for the watermark is RWM = number of luminance pixels per second cr (13) Increase of chip-rate cr, average amplitude mean( i), or variance of the pseudo- noise signal p decrease the bit error rate; using a poor lter which does not remove the video signal from the watermarked video signal very well increases the bit error rate. Table2 gives a few examples for parameters used and the resulting bit error rates. It can be seen that bit error rates below 10?10 can easily be accomplished.... In PAGE 11: ...in the right column of Table2 . However, these estimations and experiments are valid only for embedding of watermarks into uncompressed video, and do not include the e ects of malicious attacks on the watermarks.... ..."

### Table 5: The watermarking procedure analysis under the statistical averaging attack.

"... In PAGE 7: ... In the statistical averaging attack a high number of watermarked frames are collected so as the watermark can be estimated by statistical averaging. The attack has been performed by colluding about the 70% of the available frames and the obtained results are shown in Table5 . In particular, the procedure exhibits a good performance and this is essentially due to its adaptive behavior that can balance the flnal video quality with the achieved protection level.... ..."

### Table 1: Video Watermarking Methods vs. Video Characteristics

"... In PAGE 3: ... Overall, although the aforementioned papers have dealt with problems related to video watermarking, some issues still remain. In order to give the reader a clear idea of the state of the art in video watermarking, Table1 summarizes selected techniques versus video characteristics. In the following subsections, the issues of compressed domain watermarking, real-time detection, bit-rate control, and resistance to peculiar attacks that video sequences may encounter will be discussed.... In PAGE 12: ... This is reasonable because a signal embedded in the transformed domain can be transferred to another equivalent signal in the spatial domain and watermark estimation by means of denoising [10, 23] is intuitively applied in the spatial domain. As shown in Table1 , the resistance of the video frame-dependent watermark to WEAs is the novel contribution of this paper. 3.... ..."

### Table 2: Statistics of the video traces used in the simulations.

"... In PAGE 5: ... The schemes R1 and R2 estimate the initial throughput for the Internet path as the average of the first twenty throughput samples for the host. The desired sending rate of a source is either one of the above five CBR rates, or the rate of the video traces we collected, where Table2 shows statistics of the video traces. In our first experiment, connections arrive and depart accord- ing to a Poisson distribution.... ..."

### Table 1. Statistical Estimates and Conclusions

2001

"... In PAGE 3: ... 4.2 Bootstrap estimates and inferences Table1 lists statistical estimates, with standard errors in parenthesis, for NLANR-bo1, VZL and POLY obtained using both the Bootstrap and conventional statistical methods. The latter consists of calculating the desired statistic on a single sample from a population and, except for the mean, cannot provide an estimate of error.... In PAGE 3: ... A. Sample Averages and Percentiles Row (1) of Table1 is the mean of the response sizes in bytes with the standard deviation of the sample mean in parenthesis, both calculated using the Bootstrap on the full datasets. POLY has the largest mean response size with the smallest ... In PAGE 5: ... The theoretical value is expected to be 50% for identical distributions. In row (8) of Table1 , the ASL for NLANR-bo1 is 48% because the NLANR-bo1 workload is the reference. Row (8) shows that NLANR-bo1 and the VZL possess the same response size distribution with a significance level of 15%.... ..."

Cited by 1

### Table 1. Error probability of our watermark detector and its threshold

2005

"... In PAGE 8: ... 5. Table1 shows the probability of our watermark de- tector generating an error based on the chosen threshold. 5 Simulation Results This section evaluates three feature-based watermarking methods: Kutter method [1], Bas method [8], and Our method.... ..."

Cited by 1

### Table 4. Structure of the watermarked test sequence Frame Description

"... In PAGE 3: ...1. Test Sequence Generation The algorithm shown in Table 3 is used for creating a video se- quence that represents two occurrences of a given advertisement, interleaved with two other sequences of no interest ( Table4 ). The basic watermarked video sequence is then corrupted in two differ- ent ways, to model the communication problems found in televi- sion broadcasting and in wireless channels with error correction.... ..."

### Table 1 Error probability of our watermark detector and its threshold.

### Table 3: Test probabilities obtained with the Uk statistics and using the true error rate for ve repetitions of the experiment.

"... In PAGE 13: ... Better results are obtained by considering the statistics Uk; k = 0; : : : ; 8. In Table3 test probabilities for each k and for ve repetitions of the experiment are shown. One sees that the tests based on Uk rejects the undersmoothed estimates for k large with very small test probabilities, while the oversmoothed estimates are rejected for k small.... ..."

### Table 2. Distortion False positive probability False negative probability

in higher

"... In PAGE 14: ... The false positive probability and false negative probability can be computed using the estimates of mean and variance. The resulted estimated errors are shown in Table2 . Random geometric attack performance is the worst with 2 10 89 .... In PAGE 25: ...24 Table Captions Table 1. Stirmark test results compared to commercial algorithms Table2 . False positive probability and false negative probability for various distortions of the 100 watermarked image ... ..."