### Table 1. Classical error-diffusion algorithm processing the pixels in a scanline order using Floyd-Steinberg (FS) coefficient set.

### Table 1: Cumulative mutual information between coefficient magnitude C and a linear estimator l(~ Q) composed of a subset of neighbors. Each entry gives the mutual information for a subset containing the neighbors indicated at the top of that column and all columns to the left. Notice that the local neighbors within the subband (Left and Up), the Parent, and the Cousins contribute most to the mutual information. Values are averaged over two scales (levels 1 and 2) of three training images (Lena, Boats, Baboon).

1999

"... In PAGE 8: ... Rather than exhaustively explore all possible neighbor subsets, we used a greedy algorithm to choose conditioning neighbors. Table1 shows the greedy optimal neighbor subset for the three oriented subbands. Using this analysis, and imposing causality (assuming a standard scanline ordering of the bits), we decided to include neighbors corresponding to the first four table columns when coding the horizontal and vertical bands, and the first five columns for the diagonal bands.... ..."

Cited by 129

### Table 1: Cumulative mutual information between coefficient magnitude C and a linear estimator l(~ Q) composed of a subset of neighbors. Each entry gives the mutual information for a subset containing the neighbors indicated at the top of that column and all columns to the left. Notice that the local neighbors within the subband (Left and Up), the Parent, and the Cousins contribute most to the mutual information. Values are averaged over two scales (levels 1 and 2) of three training images (Lena, Boats, Baboon).

1999

"... In PAGE 8: ... Rather than exhaustively explore all possible neighbor subsets, we used a greedy algorithm to choose conditioning neighbors. Table1 shows the greedy optimal neighbor subset for the three oriented subbands. Using this analysis, and imposing causality (assuming a standard scanline ordering of the bits), we decided to include neighbors corresponding to the first four table columns when coding the horizontal and vertical bands, and the first five columns for the diagonal bands.... ..."

Cited by 129

### Table 1: Cumulative mutual information between coefficient magnitude C and a linear estimator l(~ Q) composed of a subset of neighbors. Each entry gives the mutual information for a subset containing the neighbors indicated at the top of that column and all columns to the left. Notice that the local neighbors within the subband (Left and Up), the Parent, and the Cousins contribute most to the mutual information. Values are averaged over two scales (levels 1 and 2) of three training images (Lena, Boats, Baboon).

1999

"... In PAGE 8: ... Rather than exhaustively explore all possible neighbor subsets, we used a greedy algorithm to choose conditioning neighbors. Table1 shows the greedy optimal neighbor subset for the three oriented subbands. Using this analysis, and imposing causality (assuming a standard scanline ordering of the bits), we decided to include neighbors corresponding to the first four table columns when coding the horizontal and vertical bands, and the first five columns for the diagonal bands.... ..."

Cited by 129

### Table 3: This table shows statistics obtained from running the sample clip on the constrained and uncon- strained players using a scanline order exactly the same as the original MPEG player. The fps measure- ments were obtained by running the particular MPEG player 12 times on the MPEG clip, throwing out the low and high, and then averaging the rest. (Note: the frame rates were typically within 0.03 seconds of each other)

1996

"... In PAGE 7: ... The performance statistics are listed in Table 3. The most obvious number in Table3 is the large number of extra instructions (5.8 million) required to parse a non- slice per line video over the standard MPEG player.... ..."

Cited by 7

### Table 4: This table shows the performance obtained from vertically striping the constrained and uncon- strained players on the sample clip. These figures were obtained in the same manner as in Table 3.

1996

"... In PAGE 8: ... By simply changing the ordering of data accesses, the amount of caching improvement outweighs the need to parse the GOP twice for the constrained MPEG player. Please see Table4 . The unconstrained MPEG player, although improved over the scanline order, still suffers from a large amount of overhead occurred in the first pass.... ..."

Cited by 7

### Table 1: Cumulative mutual information between coefficient magnitude, C, and a linear combination of neighbor magnitudes, l( ~ Q). Each entry gives the mutual information for a subset containing the neighbors indicated at the top of that column and all columns to the left. Notice that the local neighbors within the subband (Left and Up), the Parent, and the Cousins contribute most to the mutual informa- tion. Values are averaged over the two finest pyramid scales of three training images (Lena, Boats, Baboon).

1999

"... In PAGE 8: ... Specifically, the set is constructed incrementally: at each step, we incorporate the remaining neighbor whose inclusion maximizes the mutual information. Table1 shows the greedy optimal neighbor subset for the three oriented subbands. Using this analysis, and imposing causality (assuming a standard scanline ordering of the coefficients), we decided to include neighbors corresponding to the first four table columns when coding the horizontal and vertical bands, and the first five columns for the diagonal bands.... ..."

Cited by 129

### Table 1: Cumulative mutual information between coefficient magnitude C and a linear estimator l( ~ Q) composed of a subset of neighbors. Each entry gives the mutual information for a subset containing the neighbors indicated at the top of that column and all columns to the left. Notice that the local neighbors within the subband (Left and Up), the Parent, and the Cousins contribute most to the mutual information. Values are averaged over two scales (levels 1 and 2) of three training images (Lena, Boats, Baboon).

1999

"... In PAGE 8: ... Specifically, the set is constructed incrementally: at each step, we incorporate the remaining neighbor whose inclusion maximizes the mutual information. Table1 shows the greedy optimal neighbor subset for the three oriented subbands. Using this analysis, and imposing causality (assuming a standard scanline ordering of the bits), we decided to include neighbors corresponding to the first four table columns when coding the horizontal and vertical bands, and the first five columns for the diagonal bands.... ..."

Cited by 129

### Table 7: Image sizes and disparity levels of the four image pairs, and running times of the five selected algorithms.

2002

"... In PAGE 48: ... Finally, we take a brief look at the efficiency of the different methods. Table7 lists the image sizes and number of disparity levels for each image pair, and running times for each of the five algorithms. Clearly, the local and scanline-based methods (SSD, DP, and SO) are quite fast, while GC and Bayesian diffusion are several orders of magnitude slower.... ..."

Cited by 406

### Table 1. Comparison of results between grids with and without diagonals. New results

1994

"... In PAGE 2: ... For two-dimensional n n meshes without diagonals 1-1 problems have been studied for more than twenty years. The so far fastest solutions for 1-1 problems and for h-h problems with small h 9 are summarized in Table1 . In that table we also present our new results on grids with diagonals and compare them with those for grids without diagonals.... ..."

Cited by 11