METHOD OF PROVIDING SPACE FILLING PATTERNS

Abstract

Methods of generating Hilbert space-filling indexes using simple bit-wise transformation are described herein. One method is based on a linear-based transformation uses bit-wise AND and XOR operations on a simple input index counter. In another method, the regular structure of the transform matrix was then used to obtain a Gray-based transform at a reduced complexity. Due to the simple and regular structure of the transformation, an efficient hardware implementation of generating indexes (addresses) along the Hilbert curve results.

Description

FIELD OF THE INVENTION

The present invention is related to a method of providing space filling patterns.

BACKGROUND OF THE INVENTION

It is known to align matrix data in different ways, in order to use the data for different purposes. One method of aligning data is to use a transform, such as a matrix transform for Hilbert space-filling curve, and once aligned in a pattern according to the transform, the data can be used for different purposes.

The Hilbert curve is a one-to-one mapping between a two-dimensional (2-D) space and one-dimensional space that preserve the locality presented in the 2-D space. A mathematical history of the Hilbert space-filling curve can be found in H. Sagan, Space-Filling Curves. Springer-Verlag, 1994, ch. 2. The clustering (locality between objects) properties of the curve are discussed in B. Moon, H. V. Jagadish, C. Faloutsos, and J. H. Slatz, “Analysis of the Clustering Properties of the Hilbert Space-Filling Curve,” IEEE Trans. Knowledge and Data Eng., vol. 13, no. 1, pp. 124-141, January/February 2001.

The Hilbert curve is useful in many applications where a locality-preserving mapping is required, such as R-tree indexing, process allocation, VLSI component layout and many more. In particular, the Hilbert curve is used in a variety of image processing algorithms, such as image compression, textural analysis, image segmentation and Holographic image sampling to name a few.

Several methods of Hilbert scanning have been proposed, among them scanning by using Tensor operations and look-up tables. These methods, however do not efficiently and automatically align matrix data from one form to another for certain transforms.

SUMMARY OF THE INVENTION

Methods of generating Hilbert space-filling indexes using simple bit-wise transformation are described herein.

One method is based on a linear-based transformation uses bit-wise AND and XOR operations on a simple input index counter.

In another method, the regular structure of the transform matrix was then used to obtain a Gray-based transform at a reduced complexity. Due to the simple and regular structure of the transformation, an efficient hardware implementation of generating indexes (addresses) along the Hilbert curve results.

BRIEF DESCRIPTION OF THE DRAWINGS

These and other aspects and features of the present invention will become apparent to those of ordinary skill in the art upon review of the following description of specific embodiments of the invention in conjunction with the accompanying figures, wherein:

FIG. 1 illustrates Hilbert Filling for N=2;

FIG. 2 illustrates Hilbert Filling for N=4;

FIG. 3 illustrates Hilbert Filling for N=6;

FIG. 4 illustrates Hilbert Filling for N=8; and

FIG. 5 illustrates Hilbert Filling for N=10.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

With respect to the Hilbert transform, in a famous 1891 paper, D. Hilbert wrote that the function required for constructing the Hilbert space-filling curve can be represented in a lucid manner, if one avails oneself of its special geometric insight. The insight Hilbert was referring to is based on the fractal structure of the filling curve, where for each quadrant of the space, for each sub-space, the sub-curve follows the same structure as the curve in the lower order space (with rotations and reflections). The present invention uses this insight to formulate a general transform for generating the Hilbert space-filling indexes. A natural extension is an K-dimensional Hilbert space-filling transformation.

Due to its locality-preserving property, the Hilbert space-filling curve is used in a variety of applications, where data mapping from multiple-dimensions to one-dimension is required while preserving data clustering. An efficient algorithm to calculate the data indexes along the curve can then be considered as an address generation unit, serving the DSP application. Provided herein are various transformations based essentially on two index transformation methods using simple bit operations. A simple counter-based Hilbert transformation is derived first. By using the regular structure of the transform matrix, a significant complexity reduction is achieved by basing the Hilbert transform on Gray indexes, and one in which all computation are preferably done over GF(2). These methods lend themselves to a simple and/or efficient hardware implementation that can be implemented on a computer that has software, firmware, or a combination of both that implement a program that contains the functions described herein.

Introduction

This document provides the matrix settings and illustration figures for the Hilbert space-filling curves.

The general transform is provided in the following equation:

out=T(*)in(+)B,   [1]

where, “in” is the address of the input index, presented as a column vector where the first row of the vector is the LSB of the address and the last row of the vector is the MSB of the address. “T” is the transform matrix, “B” is the offset column vector, (*) denotes a matrix multiply using AND and XOR operations and (+) is the XOR operation. “out” is the address of the transform index, presented as a column vector as well.

The following information is provided for an 8×8 matrix transform (N=8).

Hilbert Filling

For N=2 (4 elements), the transform parameters are:

• • T=[1,1 ; . . . 0, 1];
  • B=[0, 0];

FIG. 1 illustrates Hilbert filling for N=2. Following are the Hilbert order indexes:

• • 0, 1, 3, 2Hilbert Filling using Parameterize Matrix

Even though the Hilbert filling pattern cannot be described as a constant linear matrix transformation, there is a description using a parameterize matrix.

Following are examples of Hilbert filling descriptions for N equals 4 to 10.

Hilbert 4

For N=4 (16 elements), the input index is described with 4 bits: c1c0r1r0, where r0 is the LSB and c1 is the MSB.

• • Define:

x=c0+c1,

• • where the operator “+” denotes a XOR operation.

Then, define the following matrix:

$T=\lfloor \begin{array}{cccc}x& 1& 0& \stackrel{\_}{x}\\ 0& 0& 1& 1\\ \stackrel{\_}{x}& 1& 0& \stackrel{\_}{x}\\ 0& 0& 0& 1\end{array}\rfloor$

The Hilbert indexes are then described as the “T” transform of the input indexes:

Out=T(*)in(+)B,

• • where B is a vector of zeros.

FIG. 2 illustrates Hilbert Filling for N=4. Following are the Hilbert order indexes:

• • 0, 4, 5, 1, 2, 3, 7, 6, 10, 11, 15, 14, 13, 9, 8, 12

Hilbert 6

For N=6 (64 elements), the input index is described with 6 bits: c2c1c0r2r1r0, where r0 is the LSB and c2 is the MSB.

• • Define:

x=c0+c2,

y=c1+c2

• • where the operator “+” denotes a XOR operation.

Then, define the following matrix:

$T=\lfloor \begin{array}{cccccc}\stackrel{\_}{x+y}& 1& 0& \stackrel{\_}{x}& 0& \stackrel{\_}{y}\\ 0& 0& y& 1& 0& \stackrel{\_}{y}\\ 0& 0& 0& 0& 1& 1\\ x+y& 1& 0& \stackrel{\_}{x}& 0& \stackrel{\_}{y}\\ 0& 0& \stackrel{\_}{y}& 1& 0& \stackrel{\_}{y}\\ 0& 0& 0& 0& 0& 1\end{array}\rfloor$

• • where the operator “+” denotes a XOR operation.

The Hilbert indexes are then described as the “T” transform of the input indexes:

Out=T(*)in(+)B,

• • where B is a vector of zeros.

FIG. 3 illustrates Hilbert Filling for N=6. Following are the Hilbert order indexes:

• • 0, 1, 9, 8, 16, 24, 25, 17, 18, 26, 27, 19, 11, 10, 2, 3,
  • 4, 12, 13, 5, 6, 7, 15, 14, 22, 23, 31, 30, 29, 21, 20, 28,
  • 36, 44, 45, 37, 38, 39, 47, 46, 54, 55, 63, 62, 61, 53, 52, 60,
  • 59, 58, 50, 51, 43, 35, 34, 42, 41, 33, 32, 40, 48, 49, 57, 56

Hilbert 8

For N=8 (256 elements), the input index is described with 8 bits: c3c2c1c0r3r2r1r0, where r0 is the LSB and c3 is the MSB.

• • Define:

x=r3+r2,

y=c0+c1

z=c2+c3

• • where the operator “+” denotes a XOR operation.

Then, define the following matrix:

$T=\lfloor \begin{array}{cccccccc}x+y+z& 1& 0& \stackrel{\_}{x}& 0& \stackrel{\_}{y}& 0& \stackrel{\_}{z}\\ 0& 0& \stackrel{\_}{y+z}& 1& 0& \stackrel{\_}{y}& 0& \stackrel{\_}{z}\\ 0& 0& 0& 0& z& 1& 0& \stackrel{\_}{z}\\ 0& 0& 0& 0& 0& 0& 1& 1\\ \stackrel{\_}{x+y+z}& 1& 0& \stackrel{\_}{x}& 0& \stackrel{\_}{y}& 0& \stackrel{\_}{z}\\ 0& 0& y+z& 1& 0& \stackrel{\_}{y}& 0& \stackrel{\_}{z}\\ 0& 0& 0& 0& \stackrel{\_}{z}& 1& 0& \stackrel{\_}{z}\\ 0& 0& 0& 0& 0& 0& 0& 1\end{array}\rfloor$

• • where the operator “+” denotes a XOR operation.

The Hilbert indexes are then described as the “T” transform of the input indexes:

Out=T(*)in(+)B,

• • where B is a vector of zeros.

FIG. 4 illustrates Hilbert Filling for N=8. Following are the Hilbert order indexes:

• • 0, 16, 17, 1, 2, 3, 19, 18, 34, 35, 51, 50, 49, 33, 32, 48,
  • 64, 65, 81, 80, 96, 112, 113, 97, 98, 114, 115, 99, 83, 82, 66, 67,
  • 68, 69, 85, 84, 100, 116, 117, 101, 102, 118, 119, 103, 87, 86, 70, 71,
  • 55, 39, 38, 54, 53, 52, 36, 37, 21, 20, 4, 5, 6, 22, 23, 7,
  • 8, 9, 25, 24, 40, 56, 57, 41, 42, 58, 59, 43, 27, 26, 10, 11,
  • 12, 28, 29, 13, 14, 15, 31, 30, 46, 47, 63, 62, 61, 45, 44, 60,
  • 76, 92, 93, 77, 78, 79, 95, 94, 110, 111, 127, 126, 125, 109, 108, 124,
  • 123, 122, 106, 107, 91, 75, 74, 90, 89, 73, 72, 88, 104, 105, 121, 120,
  • 136, 137, 153, 152, 168, 184, 185, 169, 170, 186, 187, 171, 155, 154, 138, 139,
  • 140, 156, 157, 141, 142, 143, 159, 158, 174, 175, 191, 190, 189, 173, 172, 188,
  • 204, 220, 221, 205, 206, 207, 223, 222, 238, 239, 255, 254, 253, 237, 236, 252,
  • 251, 250, 234, 235, 219, 203, 202, 218, 217, 201, 200, 216, 232, 233, 249, 248,
  • 247, 231, 230, 246, 245, 244, 228, 229, 213, 212, 196, 197, 198, 214, 215, 199,
  • 183, 182, 166, 167, 151, 135, 134, 150, 149, 133, 132, 148, 164, 165, 181, 180,
  • 179, 178, 162, 163, 147, 131, 130, 146, 145, 129, 128, 144, 160, 161, 177, 176,
  • 192, 208, 209, 193, 194, 195, 211, 210, 226, 227, 243, 242, 241, 225, 224, 240

Hilbert 10

For N=10 (1024 elements), the input index is described with 10 bits: c4c3c2c1c0r4r3r2r1r0, where r0 is the LSB and c4 is the MSB.

• • Define:

x=r3+r2,

y=c0+r4

z=c2+c1

q=c3+c4

• • where the operator “+” denotes a XOR operation.

Then, define the following matrix:

$T=\lfloor \begin{array}{cccccccccc}\stackrel{\_}{x+y+z+q}& 1& 0& \stackrel{\_}{x}& 0& \stackrel{\_}{y}& 0& \stackrel{\_}{z}& 0& \stackrel{\_}{q}\\ 0& 0& y+z+q& 1& 0& \stackrel{\_}{y}& 0& \stackrel{\_}{z}& 0& \stackrel{\_}{q}\\ 0& 0& 0& 0& \stackrel{\_}{z+q}& 1& 0& \stackrel{\_}{z}& 0& \stackrel{\_}{q}\\ 0& 0& 0& 0& 0& 0& q& 1& 0& \stackrel{\_}{q}\\ 0& 0& 0& 0& 0& 0& 0& 0& 1& 1\\ x+y+z+q& 1& 0& \stackrel{\_}{x}& 0& \stackrel{\_}{y}& 0& \stackrel{\_}{z}& 0& \stackrel{\_}{q}\\ 0& 0& \stackrel{\_}{y+z+q}& 1& 0& \stackrel{\_}{y}& 0& \stackrel{\_}{z}& 0& \stackrel{\_}{q}\\ 0& 0& 0& 0& z+q& 1& 0& \stackrel{\_}{z}& 0& \stackrel{\_}{q}\\ 0& 0& 0& 0& 0& 0& \stackrel{\_}{q}& 1& 0& \stackrel{\_}{q}\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 1\end{array}\rfloor$

• • where the operator “+” denotes a XOR operation.

The Hilbert indexes are then described as the “T” transform of the input indexes:

Out=T(*)in(+)B,

• • where B is a vector of zeros.

FIG. 5 illustrates Hilbert Filling for N=10. Following are the Hilbert order indexes:

• • 0, 1, 33, 32, 64, 96, 97, 65, 66, 98, 99, 67, 35, 34, 2, 3,
  • 4, 36, 37, 5, 6, 7, 39, 38, 70, 71, 103, 102, 101, 69, 68, 100,
  • 132, 164, 165, 133, 134, 135, 167, 166, 198, 199, 231, 230, 229, 197, 196, 228,
  • 227, 226, 194, 195, 163, 131, 130, 162, 161, 129, 128, 160, 192, 193, 225, 224,
  • 256, 288, 289, 257, 258, 259, 291, 290, 322, 323, 355, 354, 353, 321, 320, 352,
  • 384, 385, 417, 416, 448, 480, 481, 449, 450, 482, 483, 451, 419, 418, 386, 387,
  • 388, 389, 421, 420, 452, 484, 485, 453, 454, 486, 487, 455, 423, 422, 390, 391,
  • 359, 327, 326, 358, 357, 356, 324, 325, 293, 292, 260, 261, 262, 294, 295, 263,
  • 264, 296, 297, 265, 266, 267, 299, 298, 330, 331, 363, 362, 361, 329, 328, 360,
  • 392, 393, 425, 424, 456, 488, 489, 457, 458, 490, 491, 459, 427, 426, 394, 395,
  • 396, 397, 429, 428, 460, 492, 493, 461, 462, 494, 495, 463, 431, 430, 398, 399,
  • 367, 335, 334, 366, 365, 364, 332, 333, 301, 300, 268, 269, 270, 302, 303, 271,
  • 239, 238, 206, 207, 175, 143, 142, 174, 173, 141, 140, 172, 204, 205, 237, 236,
  • 235, 203, 202, 234, 233,2 32, 200, 201, 169, 168, 136, 137, 138, 170, 171, 139,
  • 107, 75, 74, 106, 105, 104, 72, 73, 41, 40, 8, 9, 10, 42, 43, 11,
  • 12, 13, 45, 44, 76, 108, 109, 77, 78, 110, 111, 79, 47, 46, 14, 15,
  • 16, 48, 49, 17, 18, 19, 51, 50, 82, 83, 115, 114, 113, 81, 80, 112,
  • 144, 145, 177, 176, 208, 240, 241, 209, 210, 242, 243, 211, 179, 178, 146, 147,
  • 148, 149, 181, 180, 212, 244, 245, 213, 214, 246, 247, 215, 183, 182, 150, 151,
  • 119, 87, 86, 118, 117, 116, 84, 85, 53, 52, 20, 21, 22, 54, 55, 23,
  • 24, 25, 57, 56, 88, 120, 121, 89, 90, 122, 123, 91, 59, 58, 26, 27,
  • 28, 60, 61, 29, 30, 31, 63, 62, 94, 95, 127, 126, 125, 93, 92, 124,
  • 156, 188, 189, 157, 158, 159, 191, 190, 222, 223, 255, 254, 253, 221, 220, 252,
  • 251, 250, 218, 219, 187, 155, 154, 186, 185, 153, 152, 184, 216, 217, 249, 248,
  • 280, 281, 313, 312, 344, 376, 377, 345, 346, 378, 379, 347, 315, 314, 282, 283,
  • 284, 316, 317, 285, 286, 287, 319, 318, 350, 351, 383, 382, 381, 349, 348, 380,
  • 412, 444, 445, 413, 414, 415, 447, 446, 478, 479, 511, 510, 509, 477, 476, 508,
  • 507, 506, 474, 475, 443, 411, 410, 442, 441, 409, 408, 440, 472, 473, 505, 504,
  • 503, 471, 470, 502, 501, 500, 468, 469, 437, 436, 404, 405, 406, 438, 439, 407,
  • 375, 374, 342, 343, 311, 279, 278, 310, 309, 277, 276, 308, 340, 341, 373, 372,
  • 371, 370, 338, 339, 307, 275, 274, 306, 305, 273, 272, 304, 336, 337, 369, 368,
  • 400, 432, 433, 401, 402, 403, 435, 434, 466, 467, 499, 498, 497, 465, 464, 496,
  • 528, 560, 561, 529, 530, 531, 563, 562, 594, 595, 627, 626, 625, 593, 592, 624,
  • 656, 657, 689, 688, 720, 752, 753, 721, 722, 754, 755, 723, 691, 690, 658, 659,
  • 660, 661, 693, 692, 724, 756, 757, 725, 726, 758, 759, 727, 695, 694, 662, 663,
  • 631, 599, 598, 630, 629, 628, 596, 597, 565, 564, 532, 533, 534, 566, 567, 535,
  • 536, 537, 569, 568, 600, 632, 633, 601, 602, 634, 635, 603, 571, 570, 538, 539,
  • 540, 572, 573, 541, 542, 543, 575, 574, 606, 607, 639, 638, 637, 605, 604, 636,
  • 668, 700, 701, 669, 670, 671, 703, 702, 734, 735, 767, 766, 765, 733, 732, 764,
  • 763, 762, 730, 731, 699, 667, 666, 698, 697, 665, 664, 696, 728, 729, 761, 760,
  • 792, 793, 825, 824, 856, 888, 889, 857, 858, 890, 891, 859, 827, 826, 794, 795,
  • 796, 828, 829, 797, 798, 799, 831, 830, 862, 863, 895, 894, 893, 861, 860, 892,
  • 924, 956, 957, 925, 926, 927, 959, 958, 990, 991, 1023, 1022, 1021, 989, 988, 1020,
  • 1019, 1018, 986, 987, 955, 923, 922, 954, 953, 921, 920, 952, 984, 985, 1017, 1016,
  • 1015, 983, 982, 1014, 1013, 1012, 980, 981, 949, 948, 916, 917, 918, 950, 951, 919,
  • 887, 886, 854, 855, 823, 791, 790, 822, 821, 789, 788, 820, 852, 853, 885, 884,
  • 883, 882, 850, 851, 819, 787, 786, 818, 817, 785, 784, 816, 848, 849, 881, 880,
  • 912, 944, 945, 913, 914, 915, 947, 946, 978, 979, 1011, 1010, 1009, 977, 976, 1008,
  • 1007, 1006, 974, 975, 943, 911, 910, 942, 941, 909, 908, 940, 972, 973, 1005, 1004,
  • 1003, 971, 970, 1002, 1001, 1000, 968, 969, 937, 936, 904, 905, 906, 938, 939, 907,
  • 875, 843, 842, 874, 873, 872, 840, 841, 809, 808, 776, 777, 778, 810, 811, 779,
  • 780, 781, 813, 812, 844, 876, 877, 845, 846, 878, 879, 847, 815, 814, 782, 783,
  • 751, 719, 718, 750, 749, 748, 716, 717, 685, 684, 652, 653, 654, 686, 687, 655,
  • 623, 622, 590, 591, 559, 527, 526, 558, 557, 525, 524, 556, 588, 589, 621, 620,
  • 619, 618, 586, 587, 555, 523, 522, 554, 553, 521, 520, 552, 584, 585, 617, 616,
  • 648, 680, 681, 649, 650, 651, 683, 682, 714, 715, 747, 746, 745, 713, 712, 744,
  • 743, 711, 710, 742, 741, 740, 708, 709, 677, 676, 644, 645, 646, 678, 679, 647,
  • 615, 614, 582, 583, 551, 519, 518, 550, 549, 517, 516, 548, 580, 581, 613, 612,
  • 611, 610, 578, 579, 547, 515, 514, 546, 545, 513, 512, 544, 576, 577, 609, 608,
  • 640, 672, 673, 641, 642, 643, 675, 674, 706, 707, 739, 738, 737, 705, 704, 736,
  • 768, 769, 801, 800, 832, 864, 865, 833, 834, 866, 867, 835, 803, 802, 770, 771,
  • 772, 804, 805, 773, 774, 775, 807, 806, 838, 839, 871, 870, 869, 837, 836, 868,
  • 900, 932, 933, 901, 902, 903, 935, 934, 966, 967, 999, 998, 997, 965, 964, 996,
  • 995, 994, 962, 963, 931, 899, 898, 930, 929, 897, 896, 928, 960, 961, 993, 992

Hilbert Transform Matrix Property

This section describes the property of the general Hilbert transform matrix T and presents a method to construct the matrix for all N.

Given N as the Hilbert matrix size (as shown in the previous sections), we will denote the input index as a string of binary digits:

I _l =[i _N−1 , i _N−2 , . . . i ₀], for l=0, . . . 2^N−1,   [2]

• • where i_k⊂{0,1} is a binary digit, i_N−1 is the MSB and i₀ is the LSB.
  • Define digit sum as:

$\begin{array}{cc}{s}_n=i_{2\left(n+1\right)}+i_{2\left(n+1\right)+1},\mathrm{for}n=0,\dots K\mathrm{where}K=\frac{N}{2}-2.& \left[3\right]\end{array}$

• • Define the following partial sum:

$\begin{array}{cc}{\sigma }_l=\sum _{n=l}^Ks_n,\mathrm{for}l=0,\dots K& \left[4\right]\end{array}$

Then, the structure of the Hilbert transform matrix T is as follows:

$\begin{array}{cc}{T}_{N\times N}=\lfloor \begin{array}{cccccccccccc}{\sigma }_0& 1& 0& {\stackrel{-}{s}}_0& 0& {\stackrel{-}{s}}_1& 0& {\stackrel{-}{s}}_2& 0& \dots & 0& {\stackrel{\_}{s}}_K\\ 0& 0& {\stackrel{\_}{\sigma }}_1& 1& 0& {\stackrel{-}{s}}_1& 0& {\stackrel{-}{s}}_2& 0& \dots & 0& {\stackrel{\_}{s}}_K\\ 0& 0& 0& 0& {\sigma }_2& 1& 0& {\stackrel{-}{s}}_2& 0& \dots & 0& {\stackrel{\_}{s}}_K\\ 0& 0& 0& 0& 0& 0& {\stackrel{\_}{\sigma }}_3& 1& 0& \dots & 0& {\stackrel{\_}{s}}_K\\ \dots & \dots & \dots & \dots & \dots & \dots & \dots & \dots & \dots & \dots & 0& {\stackrel{\_}{s}}_K\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& \dots & 1& 1\\ {\stackrel{\_}{\sigma }}_0& 1& 0& {\stackrel{-}{s}}_0& 0& {\stackrel{-}{s}}_1& 0& {\stackrel{-}{s}}_2& 0& \dots & 0& {\stackrel{\_}{s}}_K\\ 0& 0& {\sigma }_1& 1& 0& {\stackrel{-}{s}}_1& 0& {\stackrel{-}{s}}_2& 0& \dots & 0& {\stackrel{\_}{s}}_K\\ 0& 0& 0& 0& {\stackrel{\_}{\sigma }}_2& 1& 0& {\stackrel{-}{s}}_2& 0& \dots & 0& {\stackrel{\_}{s}}_K\\ 0& 0& 0& 0& 0& 0& {\sigma }_3& 1& 0& \dots & 0& {\stackrel{\_}{s}}_K\\ \dots & \dots & \dots & \dots & \dots & \dots & \dots & \dots & \dots & \dots & 0& {\stackrel{\_}{s}}_K\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& \dots & 0& 1\end{array}\rfloor & \left[5\right]\end{array}$

The structure of the matrix in [5] is for even N/2. For odd N/2, the first non-zero element in each row (which is the σ element) is inverted. So, for example, for N=10, the first element in the first row is σ₀.

As can be notice, the matrix has the following properties:

The N/2 bottom rows are identical to the upper N/2 rows except that the first non-zero element in each N/2 bottom row is negated relative to its counterpart (N/2 rows above it).

The first non-zero element is a partial sum. First row starts with first partial sum (σ₀), second row starts with 2 zeros follows by the second partial sum, and so on. The number of preceding zeros in each row is 2*(m−1), where ‘m’ is the row number (starting from 1).

Following the partial sum is a 1 followed by a pattern of zero and digit sum, where the digit sum starts from (m−1), where ‘m’ is the row number. Row N/2 is equal to N−2 zeros followed by two ‘1’s.

Thus, given N, construct the Hilbert transform matrix T as follows:

• • 1. Set I=0 (first input index).
  • 2. Compute digit sums using [3].
  • 3. Compute partial sums using [4].
  • 4. Construct the transform matrix using [5].
  • 5. Compute output index using [1], where B=0 and ‘in’=I.
  • 6. Set I=I+1 (next input index). If I==N stop, otherwise go to 2.

In other words, given an input index to the transform operation, the transform matrix is constructed as a function of the input index digits (digit sum and partial sum as shown in [3] and [4]). This matrix is then used to transform the input index to a Hilbert index following the linear transform equation [1].

Gray Based Space Filling

Space filling based on Gray indexing benefit from the fact that only one bit is changing between two consecutive Gray indexes. The Gray indexes serve as the source indexes for the space filling transform operation (as oppose to linear indexes, which were considered in the previous sections). The motivation for using a Gray-based index (having only one bit change between two consecutive Gray indexes) is to reduce the computational complexity.

In general, the space filling transform is defined as (we are omitting here the addition of an offset vector):

p _n =Tk _n   [6]

• • where T is the transform matrix, k is the input index (input address bits) and p is the output index (output address bits).

A Gray mapping is then defined as follows:

g _n =GK _n   [7]

• • and its inverse mapping as,

k _n =G ⁻¹ g _n.   [8]

• • where G is the Gray mapping transform.

Using the Gray indexes as the base indexes, the general space filling transform is then equal to:

p _n =T(G⁻¹g_n)=(TG⁻¹)g_n≡T_Gg_n   [9]

For N=8, the Gray mapping matrix is:

$\begin{array}{ccccccccc}G=[1,& 1,& 0,& 0,& 0,& 0,& 0,& 0;\dots & \\ & 0,& 1,& 1,& 0,& 0,& 0,& 0,& 0;\dots \\ & 0,& 0,& 1,& 1,& 0,& 0,& 0,& 0;\dots \\ & 0,& 0,& 0,& 1,& 1,& 0,& 0,& 0;\dots \\ & 0,& 0,& 0,& 0,& 1,& 1,& 0,& 0;\dots \\ & 0,& 0,& 0,& 0,& 0,& 1,& 1,& 0;\dots \\ & 0,& 0,& 0,& 0,& 0,& 0,& 1,& 1;\dots \\ & 0,& 0,& 0,& 0,& 0,& 0,& 0,& 1];\end{array}\hspace{1em}$

and the inverse-Gray mapping matrix is:

$\begin{array}{ccccccccc}{G}^{-1}=[1,& 1,& 1,& 1,& 1,& 1,& 1,& 1;\dots & \\ & 0,& 1,& 1,& 1,& 1,& 1,& 1,& 1;\dots \\ & 0,& 0,& 1,& 1,& 1,& 1,& 1,& 1;\dots \\ & 0,& 0,& 0,& 1,& 1,& 1,& 1,& 1;\dots \\ & 0,& 0,& 0,& 0,& 1,& 1,& 1,& 1;\dots \\ & 0,& 0,& 0,& 0,& 0,& 1,& 1,& 1;\dots \\ & 0,& 0,& 0,& 0,& 0,& 0,& 1,& 1;\dots \\ & 0,& 0,& 0,& 0,& 0,& 0,& 0,& 1];\end{array}\hspace{1em}$

From [9], we can write:

p _n +p _n−1 =T _G g _n+1 =T _G(g_n+g_n+1)   [10]

where the operator “+” denotes a XOR operator. Since “g” is the Gray index, the result of the XOR operation between g_n and g_n+1 is a vector with only one value of “1”. The final result is then one column of the T_G, where the column number is determine by the location of the “1” value.

We will denote d_n as the location of the value “1” in the result of the XOR operation between g_n and g_n+1:

d _n≡location of 1 in (g_n+g_n+1)   [11]

Then,

p _n+1 =p _n +T _G[d_n]  [12]

• • where T_G[d_n] denote column d_n of matrix T_G.

To summarize, the Gray-based space-filling algorithm steps are:

• • generate d_n for all n (1 to 2̂N).
  • Calculate the matrix T_G≡TG⁻¹.
  • For n=1, set g₁ to a vector of N zeros (since the first Gray index is equal to 0).
  • For each n>1, use equation [12] to calculate the output index.

Since this algorithm does not require matrix multiplication, it is less than the linear based space filling.

Gray Based Hilbert Space-Filling

As was shown in section 0, the Hilbert space filling transform matrix a function of the index number (see for example section 0 for N=8).

As such, equation [10] becomes:

$\begin{array}{cc}\begin{array}{c}{p}_n+p_{n+1}=T_{G,n}g_n+T_{G,n+1}g_{n+1}\\ =T_{G,n+1}g_{n+1}+\left(T_{G,n+1}+T_{G,n+1}+T_{G,n}\right)g_n=\\ =T_{G,n+1}\left(g_n+g_{n+1}\right)+\left(T_{G,n+1}+T_{G,n}\right)g_n\\ \equiv T_{G,n+1}\left(g_n+g_{n+1}\right)+T_{G,n,n+1}g_n\end{array}& \left[13\right]\\ \mathrm{Thus},& \\ p_{n+1}=p_n+T_{G,n+1}\left[d_n\right]+T_{G,n,n+1}g_n& \left[14\right]\end{array}$

where p₁=0 (the first Hilbert index is 0).

The matrix T_G,n+1 is the result of multiplying the Hilbert transform matrix (denoted as matrix T is section 0) with the inverse-Gray matrix. For N=8, this matrix is to:

$\begin{array}{cc}{T}_{G,n+1}=\lfloor \begin{array}{cccccccc}x+y+z& \stackrel{\_}{x+y+z}& \stackrel{\_}{x+y+z}& y+z& y+z& \stackrel{-}{z}& \stackrel{-}{z}& 0\\ 0& 0& \stackrel{\_}{y+z}& y+z& y+z& \stackrel{-}{z}& \stackrel{-}{z}& 0\\ 0& 0& 0& 0& z& \stackrel{-}{z}& \stackrel{-}{z}& 0\\ 0& 0& 0& 0& 0& 0& 1& 0\\ \stackrel{\_}{x+y+z}& x+y+z& x+y+z& \stackrel{\_}{y+z}& \stackrel{\_}{y+z}& z& z& 1\\ 0& 0& y+z& \stackrel{\_}{y+z}& \stackrel{\_}{y+z}& z& z& 1\\ 0& 0& 0& 0& \stackrel{-}{z}& z& z& 1\\ 0& 0& 0& 0& 0& 0& 0& 1\end{array}\rfloor & \left[15\right]\end{array}$

where x, y and z are defined in section 0 and the index n+1 is omitted from the matrix elements.

In the general case, the structure of the matrix T_G,n+1, is as follows:

$\begin{array}{cc}{T}_{G,n+1}=\lfloor \begin{array}{ccccccccccc}{\sigma }_0& {\stackrel{\_}{\sigma }}_0& {\stackrel{\_}{\sigma }}_0& {\sigma }_1& {\sigma }_1& \dots & {\sigma }_{K-1}& {\sigma }_{K-1}& {\stackrel{\_}{\sigma }}_K& {\stackrel{\_}{\sigma }}_K& 0\\ 0& 0& {\stackrel{\_}{\sigma }}_1& {\sigma }_1& {\sigma }_1& \dots & {\sigma }_{K-1}& {\sigma }_{K-1}& {\stackrel{\_}{\sigma }}_K& {\stackrel{\_}{\sigma }}_K& 0\\ \dots & \dots & \dots & \dots & \dots & \dots & \dots & \dots & \dots & \dots & 0\\ 0& 0& 0& 0& 0& \dots & 0& {\sigma }_K& {\stackrel{\_}{\sigma }}_K& {\stackrel{\_}{\sigma }}_K& 0\\ 0& 0& 0& 0& 0& \dots & 0& 0& 0& 1& 0\\ {\stackrel{\_}{\sigma }}_0& {\sigma }_0& {\sigma }_0& {\stackrel{\_}{\sigma }}_1& {\stackrel{\_}{\sigma }}_1& \dots & {\stackrel{\_}{\sigma }}_{K-1}& {\stackrel{\_}{\sigma }}_{K-1}& {\sigma }_K& {\sigma }_K& 1\\ 0& 0& {\sigma }_1& {\stackrel{\_}{\sigma }}_1& {\stackrel{\_}{\sigma }}_1& \dots & {\stackrel{\_}{\sigma }}_{K-1}& {\stackrel{\_}{\sigma }}_{K-1}& {\sigma }_K& {\sigma }_K& 1\\ \dots & \dots & \dots & \dots & \dots & \dots & \dots & \dots & \dots & \dots & 1\\ 0& 0& 0& 0& 0& \dots & 0& {\stackrel{-}{\sigma }}_K& {\sigma }_K& {\sigma }_K& 1\\ 0& 0& 0& 0& 0& \dots & 0& 0& 0& 0& 1\end{array}\rfloor & \left[16\right]\end{array}$

where σ_l is the partial sum defined in [4].

As can be notice, the matrix has the following properties:

• • 1. Matrix elements are constructed from the partial sum (see [4]), ‘1’s and ‘0’s.
  • 2. The bottom N/2 rows are the inverse of the upper N/2 rows for all σ_l elements (and the last column).
  • 3. Each row contains an alternative pair of σ_l (starting from column N−1 in decreasing order), where the inverse operation alternates between pairs. The last pair (in the decreasing order) duplicates one of its element one more time with an inverse operation.

The structure of the matrix in [16] is for even N/2. For odd N/2, the a elements at the beginning of each row will be inverted. So, for example, for N=6, the first row in [16] is: [ σ₀ σ₀ σ₀ σ₁ σ₁ 0]. Note that if one builds the matrix starting from column N in decreasing order, this behavior follows naturally.

We should remember that the variables σ_l (defined in [4]) are a function of an input linear index. As this section deals with Gray input index, this may present a problem. In general, the linear index is a simple counter index (increases by one for each output index), which means that it can also be used here (in addition to the input gray index, as an internal counter). However, if one prefers to use only Gray indexes, then the following can be used.

Due to the nature of the inverse Gray matrix (refer to section 0 for N=8 example of the inverse-Gray matrix transform), the following relationship between the linear index ([2]) bit and the Gray index bit can be derived:

$\begin{array}{cc}{i}_n=\sum _{l=n}^{N-1}g_l& \left[17\right]\end{array}$

where the summation label refer to XOR operations, ‘i’ refer to the linear index bits and ‘g’ refer to the Gray index bits.

Thus, following [3],

$\begin{array}{cc}{s}_n=i_{2\left(n+1\right)}+i_{2\left(n+1\right)+1}=\sum _{l=2\left(n+1\right)}^{N-1}g_l+\sum _{l=2\left(n+1\right)+1}^{N-1}g_l=g_{2\left(n+1\right)}\mathrm{for}n=0,\dots ,K& \left[18\right]\end{array}$

where, again, all operations are XOR operations. Equation [18] basically means that the digit sum is equal to the corresponding Gray index bit. Thus, using [18], the partial sum ([4]) can be derived as a function of the Gray indexes.

The matrix T_G,n,n+1 is the result of the XOR operation between T_G,n and T_G,n+1. For N=8, it is equal to:

$\begin{array}{cc}{T}_{G,n,n+1}=\lfloor \begin{array}{cccccccc}A& A& A& B& B& C& C& 0\\ 0& 0& B& B& B& C& C& 0\\ 0& 0& 0& 0& C& C& C& 0\\ 0& 0& 0& 0& 0& 0& 0& 0\\ A& A& A& B& B& C& C& 0\\ 0& 0& B& B& B& C& C& 0\\ 0& 0& 0& 0& C& C& C& 0\\ 0& 0& 0& 0& 0& 0& 0& 0\end{array}\rfloor & \left[19\right]\\ \mathrm{Where},& \\ A\equiv \left(x_n+x_{n+1}\right)+\left(y_n+y_{n+1}\right)+\left(z_n+z_{n+1}\right)B\equiv \left(y_n+y_{n+1}\right)+\left(z_n+z_{n+1}\right)C\equiv \left(z_n+z_{n+1}\right)& \left[20\right]\end{array}$

As can be notice, the matrix T_G,n,n+1 has the following attributes:

The matrix is symmetric about the rows (rows 1 to 4 are equal to rows 5 to 8).

The 4'th and 8'th output are always 0.

This means that only the first 3 rows of the matrix are required to perform multiplication.

Furthermore, the values of A, B and C can be derived as follows (for N=8):

$\begin{array}{cc}A=B=C=0;\mathrm{if}(d_n==3)A=1;\mathrm{else}\mathrm{if}(d_n==5)\{A=1;B=1;\}\mathrm{else}\mathrm{if}(d_n==7)\{A=1;B=1;C=1;\}& \left[21\right]\end{array}$

As such, there are only 4 possibilities for matrix T_G,n,n+1.

A=B=C=1. This is used only twice for N=8 (to calculate the 65'th and 193'rd outputs).

A=B=1 and C=0. This is used eight times for N=8 (to calculate the following outputs: 17'th, 49'th, 81'th, 113'th, 145'th, 177'th, 209'th and 241'st).

A=1 and B=C=0. This is used 32 time for N=8 (to calculate the following outputs: 5'th, 13'th, 21'st, 29'th, 37'th, 45'th, 53'rd, 61'st, 69'th, 77'th, 85'th, 93'rd, 101'st, 109'th, 117'th, 125'th, 133'rd, 141'st, 149'th, 157'th, 165'th, 173'rd, 181'st, 189'th, 197'th, 205'th, 213'th, 221'st, 229'th, 237'th, 245'th and 253'rd).

A=B=C=0. In this case the matrix is equal to zero and no computation is needed. This is used in most of the cases (256−2−8−32=214 times).

Note also that there is no need to keep explicitly the values of g_n for all n to compute equation [14]. Since d_n holds the position of “1” in the result of g_n+g_n+1, the value g_n can be computed recursively as follows:

g _n =g _n−1+2^(dn−1−1)   [22]

where g₁=0 (first Gray index).

Properties of T_G,n,n+1g_n

The computation of T_G,n,n+1g_n is needed only when the matrix elements are different than zero, which means if A=1, or A=B=1 or A=B=C=1. Appendix A provides the computation outputs information.

As can be notice from the Appendix, the following rule can be used to construct the result of T_G,n,n+1g_n, without performing the matrix multiply operation:

$\begin{array}{cc}\mathrm{if}\left(\left(d_n==3\right)\&\&\left(g_n\left[3\right]==0\right)\right)T_{G,n,n+1}g_n=\left[10001000\right];\mathrm{elseif}\left(\left(d_n==5\right)\&\&\left(g_n\left[5\right]==0\right)\right)T_{G,n,n+1}g_n=\left[11001100\right];\mathrm{elseif}\left(\left(d_n==7\right)\&\&\left(g_n\left[7\right]==0\right)\right)T_{G,n,n+1}g_n=\left[11101110\right];\mathrm{else}T_{G,n,n+1}g_n=\left[00000000\right];& \left[23\right]\end{array}$

Alternatively, a toggle operation can be used for each value of d_n in the {3, 5, 7} range. For example, when d_n is equal to 3, then for each g_n, the output of T_G,n,n+1g_n operation will toggle between [10001000] and [00000000] values.

Another method to generate the output of T_G,n,n+1g_n operation is as follows:

$\begin{array}{cc}\mathrm{if}\left(\left(d_n==3\right)\left(d_n==5\right)\left(d_n==7\right)\right)\{\mathrm{if}\left(g_n\left[d_n\right]==0\right)\{s=2^{\mathrm{floor}\left(d_n/2\right)}-1//s\mathrm{contains}4\mathrm{bits}T_{G,n,n+1}g_n=\left[s,s\right]\}\}\mathrm{else}T_{G,n,n+1}g_n=\left[00000000\right]& \left[24\right]\end{array}$

where the “>>” operation denotes a shift to the right operation (with truncation), and “floor” operator perform truncation.

For the general case (N=2̂k), to calculate T_G,n,n+1g_n, we have:

$\begin{array}{cc}\mathrm{if}\left(\left(d_n==3,5,...N-1\right\}\right)//\mathrm{either}\mathrm{one}\{\mathrm{if}\left(g_n\left[d_n\right]==0\right)\{s=2^{\mathrm{floor}\left(d_n/2\right)}-1//s\mathrm{contains}N\text{/}2\mathrm{bits}T_{G,n,n+1}g_n=\left[s,s\right]\}\}\mathrm{else}T_{G,n,n+1}g_n=\mathrm{vector\_of}\mathrm{\_N}\mathrm{\_zeros}& \left[25\right]\end{array}$

To summarize, the Hilbert Gray-based space-filling algorithm steps are:

• • 1. Generate d_n for all n (0 to 2̂N−1).
  • 2. For each n>0, calculate the matrix T_G,n+1 using equation [16]. Select column d_n from the matrix.
  • 3. Calculate T_G,n,n+1g_n using [25].
  • 4. Calculate the output index using equation [14].

In other words, the Hilbert Gray-based algorithm follows the general steps of:

• • 1. Generating the differential information between two consecutive Gray indexes.
  • 2. Select the appropriate column in the combined T_G,n+1=TG⁻¹ matrix.
  • 3. Calculating the Gray index transform T_G,n,n+1g_n using simple shift operation.
  • 4. Generating the Hilbert output index using steps 2 and 3.Appendix A-T_G,n,n+1g_n Behavior

The following are the T_G,n,n+1g_n input and output operands, for N=8, where: ‘count1’ denotes ‘n+1’, ‘diff_g(n)’ is d_n, ‘g(n)’ is g_n and ‘out(n+1)’ is the result of T_G,n,n+1g_n. The bit values (written inside a square bracket) are written left to right (LSB is the left most bit and MSB is the right most bit).

• • A=1, B=1, C=1
  • count1(=n+1)=65, diff_g(n)=7: g(n)=32 [0 0 0 0 0 1 0 0], out(n+1)=119 [1 1 1 0 1 1 1 0]
  • count1(=n+1)=193, diff_g(n)=7: g(n)=224 [0 0 0 0 0 1 1 1], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • A=1, B=1, C=0
  • count1(=n+1)=17, diff_g(n)=5: g(n)=8 [0 0 0 1 0 0 0 0], out(n+1)=51 [1 1 0 0 1 1 0 0]
  • count1(=n+1)=49, diff_g(n)=5: g(n)=56 [0 0 0 1 1 1 0 0], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=81, diff_g(n)=5: g(n)=104 [0 0 0 1 0 1 1 0], out(n+1)=51 [1 1 0 0 1 1 0 0]
  • count1(=n+1)=113, diff_g(n)=5: g(n)=88 [0 0 0 1 1 0 1 0], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=145, diff_g(n)=5: g(n)=200 [0 0 0 1 0 0 1 1], out(n+1)=51 [1 1 0 0 1 1 0 0]
  • count1(=n+1)=177, diff_g(n)=5: g(n)=248 [0 0 0 1 1 1 1 1], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=209, diff_g(n)=5: g(n)=168 [0 0 0 1 0 1 0 1], out(n+1)51 [1 1 0 0 1 1 0 0]
  • count1(=n+1)=241, diff_g(n)=5: g(n)=152 [0 0 0 1 1 0 0 1], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • A=1, B=0, C=0
  • count1(=n+1)=5, diff_g(n)=3: g(n)=2 [0 1 0 0 0 0 0 0], out(n+1)=17 [1 0 0 0 1 0 0 0]
  • count1(=n+1)=13, diff_g(n)=3: g(n)=14 [0 1 1 1 0 0 0 0], out(n+1 1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=21, diff_g(n)=3: g(n)=26 [0 1 0 1 1 0 0 0], out(n+1)=17 [1 0 0 0 1 0 0 0 0]
  • count1(=n+1)=29, diff_g(n)=3: g(n)=22 [0 1 1 0 1 0 0 0], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=37, diff_g(n)=3: g(n)=50 [0 1 0 0 1 1 0 0], out(n+1)=17 [1 0 0 0 1 0 0 0]
  • count1(=n+1)=45, diff_g(n)=3: g(n)=62 [0 1 1 1 1 1 0 0], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=53, diff_g(n)=3: g(n)=42 [0 1 0 1 0 1 0 0], out(n+1)=17 [1 0 0 0 1 0 0 0]
  • count1(=n+1)=61, diff_g(n)=3: g(n)=38 [0 1 1 0 0 1 0 0], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=69, diff_g(n)=3: g(n)=98 [0 1 0 0 0 1 1 0], out(n+1)=17 [1 0 0 0 1 0 0 0]
  • count1(=n+1)=77, diff_g(n)=3: g(n)=110 [0 1 1 1 0 1 1 0], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=85, diff_g(n)=3: g(n)=122 [0 1 0 1 1 1 1 0], out(n+1)=17 [1 0 0 0 1 0 0 0]
  • count1(=n+1)=93, diff_g(n)=3: g(n)=118 [0 1 1 0 1 1 1 0], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=101, diff_g(n)=3: g(n)=82 [0 1 0 0 1 0 1 0], out(n+1)=17 [1 0 0 0 1 0 0 0]
  • count1(=n+1)=109, diff_g(n)=3: g(n)=94 [0 1 1 1 1 0 1 0], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=117, diff_g(n)=3: g(n)=74 [0 1 0 1 0 0 1 0], out(n+1)=17 [1 0 0 0 1 0 0 0]
  • count1(=n+1)=125, diff_g(n)=3: g(n)=70 [0 1 1 0 0 0 1 0], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=133, diff_g(n)=3: g(n)=194 [0 1 0 0 0 0 1 1], out(n+1)=17 [1 0 0 0 1 0 0 0]
  • count1(=n+1)=141, diff_g(n)=3: g(n)=206 [0 1 1 1 0 0 1 1], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=149, diff_g(n)=3: g(n)=218 [0 1 0 1 1 0 1 1], out(n+1)=17 [1 0 0 0 1 0 0 0]
  • count1(=n+1)=157, diff_g(n)=3: g(n)=214 [0 1 1 0 1 0 1 1], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+10=165, diff_g(n)=3: g(n)=242 [0 1 0 0 1 1 1 1], out(n+1)=17 [1 0 0 0 1 0 0 0]
  • count1(=n+1)=173, diff_g(n)=3: g(n)=254 [0 1 1 1 1 1 1 1], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=181, diff_g(n)=3: g(n)=234 [0 1 0 1 0 1 1 1], out(n+1)=17 [1 0 0 0 1 0 0 0]
  • count1(=n+1)=189, diff_g(n)=3: g(n)=230 [0 1 1 0 0 1 1 1], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=197, diff_g(n)=3: g(n)=162 [0 1 0 0 0 1 0 1], out(n+1)=17 [1 0 0 0 1 0 0 0]
  • count1(=n+1)=205, diff_g(n)=3: g(n)=174 [0 1 1 1 0 1 0 1], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=213, diff_g(n)=3: g(n)=186 [0 1 0 1 1 1 0 1], out(n+1)=17 [1 0 0 0 1 0 0 0]
  • count1(=n+1)=221, diff_g(n)=3: g(n)=182 [0 1 1 0 1 1 0 1], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=229, diff_g(n)=3: g(n)=146 [0 1 0 0 1 0 0 1], out(n+1)=17 [1 0 0 0 1 0 0 0]
  • count1(=n+1)=237, diff_g(n)=3: g(n)=158 [0 1 1 1 1 0 0 1], out(n+1)=0 [0 0 0 0 0 0 0 0]
  • count1(=n+1)=245, diff_g(n)=3: g(n)=138 [0 1 0 1 0 0 0 1], out(n+1)=17 [1 0 0 0 1 0 0 0]
  • count1(=n+1)=253, diff_g(n)=3: g(n)=134 [0 1 1 0 0 0 0 1], out(n+1)=0 [0 0 0 0 0 0 0 0]

Although the present invention has been particularly described with reference to embodiments thereof, it should be readily apparent to those of ordinary skill in the art that various changes, modifications and substitutes are intended within the form and details thereof, without departing from the spirit and scope of the invention. Accordingly, it will be appreciated that in numerous instances some features of the invention will be employed without a corresponding use of other features. Further, those skilled in the art will understand that variations can be made in the number and arrangement of components illustrated in the above figures. It is intended that the scope of the appended claims include such changes and modifications.

Claims

1. A method of obtaining a Hilbert index from a linear-based index, the method comprising the step of: performing a bit-wise transformation on the linear-based index to obtain the Hilbert index.

2. The method according to claim 1, wherein the step of performing includes the steps of: selecting an input index length that corresponds to the linear-based index, where the input index length is an even integer greater than or equal to 2;constructing a Hilbert transform matrix for each linear-based index digit, wherein each Hilbert transform matrix is filled using digit sums of the associated linear-based index digit, partial sums of the associated linear-based index digit, zeros and ones; andmultiplying each linear-based index digit with the associated Hilbert transform matrix to obtain the Hilbert index that contains the associated Hilbert index digits.

3. The method according to claim 1 wherein the step of constructing each Hilbert transform matrix includes the steps of: setting the input index I to 0;computing digit sums of the set input index using equation

4. The method according to claim 2 wherein the steps of selecting, constructing and multiplying are automatically performed.

5. The method according to claim 3 wherein a processor executes a software program to automatically perform the steps of selecting, constructing and multiplying.

6. The method according to claim 3 wherein the steps of selecting, constructing and multiplying are automatically performed using at least some logic circuitry to obtain the Hilbert index.

7. The method according to claim 5 wherein the steps of selecting, constructing and multiplying are automatically performed entirely using logic circuitry to obtain the Hilbert index.

8. The method according to claim 1 wherein each of the associated Hilbert index digits are precalculated and stored in a memory as precalculated Hilbert index digits; and further including the step of: receiving a particular linear based index digit; andretrieving from the memory, using the particular linear based index digit, an associated one of the precalculated Hilbert index digits.

8. A method of obtaining a Hilbert index from a Gray-based index, the method comprising the step of: performing a bit-wise transformation on the Gray-based index to obtain the Hilbert index.

9. The method according to claim 8, wherein the step of performing includes the steps of: selecting an input index length that corresponds to the Gray-based index, where the input index length is an even integer greater than or equal to 2;constructing, for the selected input index length, a transform matrix that is a combination of a linear-based Hilbert matrix and an inverse Gray matrix for each Gray-based index digit; andselecting a specific column of the transform matrix for each Gray-based index digit; andoperating upon each Gray-based index digit with the associated specific column of the transform matrix, an offset variable, and a previous Hilbert output index to obtain the Hilbert index that contains the associated Hilbert index digits.

10. The method according to claim 9 wherein the step of operating includes the steps of: generating differential information data elements between two consecutive Gray-based indexes for all Gray indexes corresponding to the Gray-based index digits;for each generated differential information data, selecting the associated specific column within the transform matrix, as a function of the differential information;calculating the offset variable as a function of the differential information using shift operations; andfor each Gray-based index digit, adding the offset variable to the associated specific column and to the previous Hilbert output index to create the Hilbert index.

11. The method according to claim 10 wherein the steps of selecting, constructing and transforming are automatically performed.

12. The method according to claim 11 wherein a processor executes a software program to automatically perform the steps of selecting, constructing and transforming.

13. The method according to claim 11 wherein the steps of selecting, constructing and transforming are automatically performed using at least some logic circuitry to obtain the Hilbert index.

14. The method according to claim 11 wherein the steps of selecting, constructing and transforming are automatically performed entirely using logic circuitry to obtain the Hilbert index.

15. The method according to claim 9 wherein precalculated Hilbert indexes are retrieved based upon a previous differential information data element and a previous Hilbert index.