1686 IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 16, NO.

6, JUNE 2007

Correspondence
Reversible Integer Color Transform TABLE I
PARAMETERS OF THE REVERSIBLE INTEGER COLOR TRANSFORMS (PART 1)
Soo-Chang Pei and Jian-Jiun Ding

Abstract—In this correspondence, we introduce a systematic algorithm

that can convert any 3 3 color transform into a reversible integer-to-in-
teger transform. We also discuss the ways to improve accuracy and reduce
implementation complexity. We derive the integer RGB-to-KLA, IV V
,
YC C , DCT, YUV, and YIQ transforms that are optimal in accuracy.
Index Terms—Color image processing, color transform, integer
transform.

I. INTRODUCTION
In color image processing, there are a variety of color-coordinate
systems. Most of them consist of three components. Some popular and
well-known ones are RGB, Karhunen–Loeve average (KLA), highest
ability for color decorrelation, IV1 V2 , YCb Cr , DCT, YUV, and YIQ
[1], [2]. In many applications, we usually have to transform one color
coordinate system into another one. The transformation is done by a 3
2 3 matrix. For example, the RGB-to-KLA transform is [2]
0:54933 0:60238 0:57912
AKLA = 0:80429 00 19322 00 56194
: : : (1)
0:22661 00 77447 00 59063
: :
Since the entries of color transform matrices are usually not of binary
form, we should use the floating-point processor to implement them.
When using the fixed-point processor, we should approximate them by
binary matrices. Unfortunately, the approximate2d matrices are always
irreversible. For example, suppose that we apply rounding to approxi-
A R
mate KLA by a binary matrix, i.e., KLA = round( KLA 2k )20k A
S
and KLA = round( KLA A01 (2k )20k . From computer experiments,
we find that
SKLA 1 RKLA 6= I no matter how large k is: (2)
Thus, after approximating a color transform by sums of powers of two,
the reversibility property is always lost. It affects the performances of
many image processing applications. For example, we usually hide the
watermark in the least significant bit. If we cannot recover the original
image, even if only the least significant bit is wrong, the watermark will
be destroyed.
If we want to preserve the reversibility property, we should use
a floating-point processor, which is more time-consuming and in-
efficient. To overcome this problem, some integer color transforms
used for approximating noninteger color transforms were developed
[2]–[6], [14], [15].
Manuscript received May 4, 2006; revised January 16, 2007. This work was
supported by the National Science Council, R.O.C., under Contracts 93-2219-E-
002-004 and NSC 93-2752-E-002-006-PAE. The associate editor coordinating
the review of this paper and approving it for publication was Dr. G. Marcu.
The authors are with the Department of Electrical Engineering, National
Taiwan University, Taipei, Taiwan 10617, R.O.C. (e-mail: pei@cc.ee.ntu.
edu.tw; djj@cc.ee.ntu.edu.tw).
Digital Object Identifier 10.1109/TIP.2007.896617
II. PREVIOUS WORKS ABOUT INTEGER TRANSFORMS
AND MATRIX FACTORIZATION
In [2], the reversible integer RGB-to-KLA transform was introduced.
In [3], the reversible integer RGB-to-YCb Cr was developed. Then,
in [4]–[6], [14], Hao, Shi, and Chen used a systematic way based on
matrix factorization to derive the integer color transforms.
Matrix factorization is an important problem in linear algebra [12].
In [9]–[11], the ways that use three triangular matrices to decompose a
unitary matrix were proposed. In [13], the way to decompose a matrix
into a product of one-row matrices was also proposed. In [4]–[6], [14],
Hao, Shi, and Chen apply these matrix decomposition methods to de-
rive the reversible integer color transforms.
In this paper, we improve the previous works about integer color
transform derivation. Our algorithm is also based on matrix decompo-
sition. The improvement is that the proposed algorithm can achieve all
the following five goals at the same time.
• [Goal 1] The integer color transform should be reversible.
• [Goal 2] No floating-point processor is required for both the for-
ward and the inverse transforms.
• [Goal 3] The bit-length of the output should be constrained.
• [Goal 4] Less complexity for implementation
• [Goal 5] Accuracy: The integer transforms should well approxi-
mate the original transform.
Compared with the works in [2] and [3], the integer color transforms
derived here are much more accurate (see Section VII). Compared with
the advanced works in [4]–[6], [14], the improvements are: 1) the ways
to save the number of time cycles for implementation (Section IV) and
2) the method to analyze the accuracy by normalized root mean square
error are proposed (Section V). 3) The closed-form solutions of coef-
ficients are shown. 4) Instead of 61, the entries of diagonal matrices
can be 62k . Moreover, two diagonal matrices were applied. Since our
algorithm is more general, the integer color transform with even higher
accuracy can be obtained. 5) We use our algorithm to derive the integer

1057-7149/$25.00 © 2007 IEEE

IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 16, NO. 6, JUNE 2007 1687

TABLE II then, in sum, there are

PARAMETERS OF THE REVERSIBLE INTEGER COLOR TRANSFORMS (PART 2)
2
(3!) ( pk + 1)3 26=2 = 1152(pk + 1)3 (8)

C
choices for . In theory, we can set pk (the upper bound of km ) to
C
infinite, i.e., there are infinite choices for . However, from our exper-
iments, the probability that the optimal integer transform is obtained
when km has a large value is very small. Thus, in (7), setting pk = 2
is usually sufficient to find the optimal integer transform. In this case,
there are 31 104 choices for . C
Then, applying the lifting scheme and the single-row elementary re-
versible matrix scheme [6], [7] with several modifications, we can de-
C
compose into four one-row matrices

C T 4 T 3 T2 T1
= (9)
t1 t2
1

T1 = 0 1 0 ; where t1 = (c22 0 1)=c21

0 0 1

t2 = 0 (t1 z 2 + z1 ) (10)
z1 c21 c22 01 0c23
1 0 c33
=
z2 c31 c32
cm1n are the entries of C (11)
1 0 0

RGB-to-KLA, IV1 V2 , XYZ, UVW, YIQ, YCb Cr , YUV, RS GS BS , T2 = t3 1 t4

and RC GC BC transforms that are optimal in accuracy successfully. 0 0 1
They are listed in Tables I and II. where t3 = c21 and t4 = 0z2 (12)
1 0 0

III. DECOMPOSITION AND INTEGERIZATION T3 = 0 1 0 ; where t5 = c22 c31 0 c21 c32

t 5 t6 1
First, we normalize the original 3 2 3 color transform A0 as A such and t6 = c32 0 t1 c31 (13)
A
that det( ) = 61 1 t7 t8
T4 = 0 1 0 ; where t7 = c12 0 t1c11 0 t6 t8
A =  1 A0 :  = jdet(A0 )j01=3 : (3) 0 0 1

and t8 c13 + z1 c11 + z2 c12 :

Note that, if =y Ax y Ax y y
and 1 = 0 , then =  1 . If the most
= (14)

y y
and least significant bits of 1 are 2k and 2k , then those of 1 are
k and 2k , where k  k + log  and k  k + log  . Since
2 3 1 2 4 2 2 We then use binary values gn to approximate tn
y
k3 0 k4 + 1  k1 0 k2 + 1, the number of bits of is almost the same as
y
that of 1 . Thus, scaling just shifts the dynamic range in density values
gn = Qb (tn ); n = 1; 2; . . . ::; 8 (15)
and does not enlarge or shorten it.
Then we do permutation, scaling, and sign changing operations for
A where Qb is a truncation operation, which throws the bits that are less
than 20b
C D1P1AP2D2
= (4) 1 b
Qb 6 dn 20n = 6 dn 20n ( dn = 0 or 1): (16)
where P1 is a row-permuting matrix and P2 is a column-permuting n=01 n=01
matrix. In each row and column of a permuting matrix, only one entry
is 1 and others are 0. D1 and D2 are diagonal matrices
Then we derive the reversible integer transform B that approximates
D1 [m; m] = 6 20k A from
D2 [m; m] = 6 2k (m = 1; 2; and 3) (5)
B PT1 D1V4V3 V2 V1 D2 P2T
= (17)
D1 or 2 [m; n] = 0 when m 6= n: (6)

D1 and D2 have the effects of scaling and sign changing.

where
Note that
there are 3! choices for P1 , 3! choices for P2 , and 6 choices for the 1 g1 g2 1 0 0

V1 V2
2

signs of the diagonal entries of D1 or 2 m; m . If we do not consider

[ ] = 0 1 0 ; = g3 1 g4
the case where det(C 0 and constrain that
) = 1 0 0 1 0 0 1

1 0 0 1 g7 g8
V3 = 0 1 0 ; V4 = 0 1 0 : (18)
0  km  pk (7) g5 g6 1 0 0 1
1688 IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 16, NO. 6, JUNE 2007

Theorem 1: In (17), B
is a binary matrix that approximates . A in hardware), and multiplying 20k (just doing bit shifting) does not re-
Moreover, we can define the inverse integer transform as 0 = B quire any time cycle. Then, in (18), each of 1 , 2 , 3 , and 4 requiresV V V V
PD V V V V D P
2 2
01 01 01 01 01 01 1 . It is easy to prove that
1 2 3 4 1 three time cycles. For example, in (22), to implement 1 , we must cal- V
(1) culate l1 = g1 x1 [2] in the first time cycle, calculate l1 = l1 + g2 x1 [3]
B0 B = I : (19) in the second time cycle, and calculate x2 [1] = x1 [1] + Qr fl1 g in the
third time cycle. Thus, we need 12 time cycles to implement the integer
(2) Since the inverse of a binary one-row matrix is also a binary color transform if we apply (17) and (18) directly.
one-row matrix However, from the facts that 1 two adjacent one-row matrix can be
implemented together and 2 if r is larger than q1 where 20q is the
1 0 g1 0 g2
least significant bit of x1 [1], then (22) is equivalent to
V101 = 0 1 0
0 0 1 x2 [1] = x1 [1] + Qr f1g x1 [2] + g2 x1 [3] g
1 0 0 = Qr f x1 [1] + g1 x1 [2] + g2 x 1 [3]g (25)
V201 = 0 3 g 1 0 g4

0 0 1 we can reduce the number of time cycles into 5, as (26)–(30).

1 0 0 The process of the forward integer color transform is as follows.
V301 = 0 1 0 Step 1)
0 0 f [1] = f [1] + g1 f [2]; f [2] = f [2]; f [3] = f [3]

1
g5

0 g7
g6

0
1
g8
where f =D P x T
2 2 : (26)
V01 =4 0 1 0 (20) Step 2)
0 0 1 f [1] = Qr0k f f [1] + g2 f [3] g
B
therefore, 0 is bound to be a binary matrix and the first two f [2] = f [2] + g4 f [3]; f [3] = f [3]: (27)
goals listed in Section II are satisfied.
Step 3)
f [1] = f [1]; f [2] = Qr 0k f f [2] + g3 f [1] g
f [3] = f [3] + g5 f [1]: (28)
IV. BIT CONSTRAINT

Then we try to satisfy Goals 3–5 in Section II. Note that, in (15)–(18), Step 4)
if the least significant bit of gn is 20b , and 20a and 20c are the least
f [1] = f [1] + g7 f [2]; f [2] = f [2]

x
significant bits of and ( = z z Bx), then f [3] = Qr0k f f [3] + g6 f [2] g : (29)

Step 5)
c = a + 4b: (21)
f [1] = Qr0k f f [1] + g8 f [3] g ; f [2] = f [2]; f [3] = f [3]

z
Thus, the bit-length of is much longer than that of . For example, if x z = P1T D1f (30)
the least significant bit of gn is 1=28 , then h2 = h1 + 32. To solve the
V V V
problem, we can convert 1 , 2 , 3 and 4 in (18) into addition- V where Qr is defined in (16) and km , gm , 1 , 2 , 1 , and 2 are D D P P
truncation operations. For example, we can convert 2 = 1 1 into x Vx defined in Section III. The process of the inverse integer color trans-
form is as follows.
x2 [1] =x1 [1] + Qr f g1 x1 [2] + g2 x1 [3] g Step 1)
f [1] = f [1] 0 g7 f [2]; f [2] = f [2]; f [3] = f [3]
x2 [2] =x1 [2]; x2 [3] = x1 [3]; Qr is defined in (16): (22)
where f = D101P1z : (31)
Note that, if the least significant bit of x1 [m]; x2 [1], and gn are 20q ,
0q , and 20b , respectively, then q2 = max[q1 ; min(r; q1 + b)]. If we Step 2)
2 f [1] = Qr0k f f [1] 0 g8 f [3] g ; f [2] = f [2]
set r to satisfy that q1  r < q1 + b and b > 0, then
f [3] = f [3] 0 g6 f [2]: (32)
q2 = r (independent of b): (23)
Step 3)
That is, the least significant bit of x2 [1] is increased from 20q 0b to
f [1] = f [1]; f [2] = f [2] 0 g3 f [1]

2
0r and the bit length is reduced. Although the bit length is reduced, f [3] = Qr0k f f [3] 0 g5 f [1] g : (33)
the reversibility property is preserved. For example, in (22), if we
x x
want to recover 1 from 2 , then we can apply
Step 4)
f [1] = f [1] 0 g2 f [3]; f [2] = Qr 0k f f [2] 0 g4 f [3] g
x1 [1] = x2 [1] 0 Qr f g1 x2 [2] + g2 x2 [3] g f [3] = f [3]: (34)

x1 [2] = x2 [2]; x1 [3] = x2 [3]: (24) Step 5)

f [1] = Qr0k f f [1] 0 g1 f [2] g ; f [2] = f [2]; f [3] = f [3]

x = P2D201x1 : (35)
V. TIME CYCLE PROBLEM AND IMPLEMENTATION
We can prove that
There is another problem for implementing the integer color trans- Theorem 2: If 20a is the least significant bit of , 20b is the least x
form. That is, too many time cycles are required. Suppose that in each significant bit of gn , and
time cycle we can only do one addition and one multiplication in each
entry. Quantization (just throwing bits), permuting (just twisting circuit a  r a + b + kn 0 km ; b > km 0 kn for any m; n; (36)
IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 16, NO. 6, JUNE 2007 1689

then the least significant bit of the output z is Theorem 3: The error of the integer color transform can be approx-
imated by [from (42) and (43)]
20r (independent of b): (37)
z 0 y  P1TD1 T4T3 T211 + P1TD1 T4 T312
Therefore, for our algorithm, if b is sufficient large such that (36) is + P1TD1 T413 + P1T D114
satisfied, then the least significant bit of the output is determined by
how many bits are preserved by the truncation operation Q and is in- + P1TD1 T4 T3 T2 r1 D2 P2Tx
dependent of b. That is, in (15), no matter how many bits we use for + P1TD1 T4 T3 r2 T1 D2 P2Tx
t
approximating n , the least significant bit of the output is remained + P1TD1 T4 r3 T2 T1 D2 P2Tx
to be .r + P1TD1 r4 T3 T2 T1 D2 P2Tx: (44)
Note that both the forward and inverse integer color transforms re-
quire only five time cycles. Moreover, only three storages are re- 1) The truncations in Steps 2)–5) cause P1 D1 T4 T3 T21 1 ,
T
[1] [2] [3]
quired during the process (for f , f , and f ). Thus, if we use P1T D1T4 T312 , P1TD1 T413 , and P1TD114 , respectively.
(26)–(35) for implementing the integer color transform, Goals 3 and 4 2) Quantizing T1 , T2 , T3 , and T4 into V1 , V2 ,
in Section II are satisfied. VT3 , and V4 cause PT1TD1 T4T3T2r1D2 P2TTx,
P1 D1T4 T3r2T1D2 P2 x, P1 D1 T4r3 T2 T1D2 P2 x,
T
and P1T D1 r4 T3 T2 T1 D2 P2T x, respectively.
There are some things to be noticed.
A) From Theorem 2, the least significant bit of z is independent of
VI. ACCURACY ANALYSIS
Then we discuss the problem of accuracy, i.e., Goal 5 in Section II. b. Thus, in (15) we can choose a large b. If b is very large, then
Note that, in Steps 2)–5), the truncation operation Qr0k is equivalent r r r r
1 , 2 , 3 , and 4 are very small and the last four terms in
to adding a small number (44) can be ignored

Qr0k fag = a + ; where 0 20r+k 01    20r+k 01 : (38) z 0 y  P1TD1 T4T3 T211 + P1TD1 T4T312
If the input of Qr0k is not known,  can be treated as a random +P1T D1 T413 + P1TD114 : (45)
variable that uniformly distributes in (0 0r+k 01 , 0r+k 01 ) and
2 2 For example, for the case of the integer KLA transform, when
r =0and b =8
, the last four terms affect only 8.05% of error.

[ ] = 0;
E [ 2 ] = 40r+k =12
E (39)
When r =0and b = 12 , these terms contribute only 0.07% of
error.
x
Note that in this case the error is independent of the input .
where E means the expected value. Thus, the process in (26)–(30) can T T T
B) From (44) and (45), we can see that the entries of 1 , 2 , 3 ,
be rewritten as T
and 4 affect the error of approximation. Especially, from (45),
z = P D1 V4 V3 V2 V1D P x + 11
T T T T T T
the entries of 3 and 4 have larger effects than 1 and 2 .
1 2 2 P P D
Thus, to make error small, we should adjust 1 , 2 , 1 , and
+112 ] + 13 g + 14 ) (40) D2 to make the values of t5 , t6 , t7 , and t8 in (13) and (14) as
where Vk is defined in (18) and 11 , 12 , 13 , and 14 are 3 2 1 random
small as possible.
C) After the integer transform is designed, we can use (44) together
vectors
with the following equation to estimate the normalized root mean
11 [2] = 11 [3] = 12 [1] = 12 [3] square error (NRMSE):

= 13 [1] = 13 [2] = 14 [2] = 14 [3] = 0 (41)

= E [(z 0 y)T (z 0 y)]
E [y T y ]
NRMSE (46)
and 11 [1], 12 [2], 13 [3], and 14 [1] are random variables
whose statistical characteristics are the same as in (39). We can
[]
where E means the expected value. In (4), we can vary the
compare (40) with the original noninteger color transform. If
y = Ax x
(i.e., the original noninteger color transform of ), then P P D D
permuting matrices 1 , 2 , 1 , and 2 iteratively and calcu-
y=P D T T T T D P x T
T T
1 1 4 3 2 1 2 2 , k is defined in (10)–(14)
late the NRMSE of the integer transform for each time. In (8),
1152( +1)
we discussed that there are pk 3 possible integer color
z 0 y = P1TD1 V4 V3 V211 + P1T D1V4 V312 transforms we can obtain. We can calculate NRMSE for each in-
+ P1TD1 V413 + P1TD114 teger color transform and choose the optimal one with the min-
imal NRMSE.
+ P1TD1 (V4 V3 V2 V1 0 T4 T3 T2 T1 )
2 D2 P2Tx: (42) VII. EXAMPLES
Suppose that b in (15) is large enough such that Vn = Tn + rn , where In this section, we show some integer color transforms we derived.
the entries of rn is very small compared with those of Tn , n = 1, 2, Suppose that the least significant bit of the input data is 1 (i.e., in The-
3, 4 orem 2, a =0 ). We choose the values of b and r as

V4 V3 V2 V1 0 T4 T3 T2 T1 b = 8; r = 0: (47)
= (T4 + r4 )(T3 + r3 )(T2 + r2 )(T1 + r1 )
0 T4 T3 T2 T1 Then, from Theorem 2, the least significant bit of the output is

 T4 T3 T2 r 1 + T4 T3 r 2 T1 + T4 r 3 T2 T1
+ r 4 T3 T2 T1 : (43) 200 = 1: (48)
1690 IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 16, NO. 6, JUNE 2007

Therefore, the least significant bits of both the input and the output data
are 1. We also suppose that the input signal is uniformly distributed in
Step 2)
f[1] = 02 f [1] + 641 [3] 256g
Q f f =

[0, 255] such that f[2] = [2] 0 77 [3] 128

f f =

[3] = [3] (56)

( [ ]) = 127 5 [ ] = 2552 3
f f :
2
E x n : ; E x n =

E
2
( [ ]) = 127 52
x n : : (49)
Step 3)
f [1] = [1] [2] = 01 f [2] 0 115 [1] 256g
f ; f Q f f =

Using (49) together with (39), (44), (46), and the fact that yT y = f [3] = [3] + 21 [1] 64
f f = : (57)
xT AT Ax, we can calculate the NRMSE. We try to convert the ten Step 4)
f [1] = [1] 0 57 [2] 256 [2] = [2]
f f = ; f f
color transforms in (50)–(54) into reversible integer transforms. The
results are shown in Table I and II, where gk k  D1 , D2 , P1 , ( = 1 8) f [3] = 0 f [3] + 73 [2] 128g
Q f f = : (58)
and P2 are the parameters, diagonal matrix, and permuting matrices
using in the processes of (26)–(30) (for the forward transform) and
Step 5)
z [1] = [2] 2 [2] = [3]
f = ; z f

(31)–(35) (for the inverse transform) z [3] = 02 f [1] + 193 [3] 256g 4
Q f f = = : (59)
(1) RGB to KLA Note that, in Tables I and II, the NRMSEs of all the integer color trans-
0 8185 0 8975 0 8629
: : : forms are smaller than 0.3%. In contrast, the NRMSE of the simple
1 1984 00 2879 00 8373
: : : integer RGB-to-KLA transform in [2] is 68.80%. The NRMSE of the
0 3376 01 1539 00 8800
: : : YC C
integer RGB-to- b r transform in [3] is 41.14%. For the integer
(2)RGB to IV1 V2 RGB-to-DCT transform in [5], the NRMSE is 0.3363% if we also
1p 1p 1p choose r =0 =8
and b . For the RGB-to-DCT transform, we derive
01 p 6 01 p6 2 6
= = = (50) NRMSE = 0 2679%
: : (60)
1 6 01 6 0
[ ]j in (5) can be nonunitary and that
= =

(3)RGB to XYZ Due to the fact that jD1 or 2 m; m

0 8706 0 2496 0 2868
: : :
we try all the 31 104 possible ways for decomposition, we can obtain
0 4288 0 8419 0 1635
: : :
more accurate integer color transforms.
0 0 0947 1 6006 : :
VIII. CONCLUSION
(4)RGB to UVW We introduced a systematic way to convert a 3 2 3 color transform
0 8387 0 2402 0 2754
: : :

0 6192 1 2156 0 2361

: : : (51)
matrix by a reversible integer transform, which can satisfy all the five
goals (binary entry, reversibility, short bit length, less time cycle, and
0 3003 1 7125 1 2984
: : : accuracy) listed in Section II at the same time. With the proposed in-
(5) RGB to YIQ teger color transforms, we can use the fixed-point processor instead
0 4722 0 9270 0 1800
: : : of the floating-point one to do lossless color coordinate transforma-
0 9412 00 4327 00 5085
: : : tion. It is useful for improving the efficiency and accuracy of image
0 3332 00 8259 0 4927
: : : processing.
(6)RGB to DCT
0 5774 0 5774 0 5774
: : :
REFERENCES
0 7071 0
: 00 7071 : (52)
[1] W. K. Pratt, Digital Image Processing, 2nd ed. New York: Wiley,

0 4082 00 8165 0 4082

: : :
1991.
[2] B. Deknuydt, J. Smolders, L. V. Eycken, and A. Oosterlinck, “Color
(7) RGB to YCb Cr space choice for nearly reversible image compression,” Proc. SPIE, vol.
0 299 0 587 0 114
: : :
1818, pp. 1300–1311, 1992.

0 500 00 419 00 081

: : :
[3] M. J. Gormish, E. L. Schwartz, and A. Keith et al., “Lossless and nearly
lossless compression for high quality images,” Proc. SPIE, vol. 3025,
00 169 00 331 0 500
: : : pp. 62–70, Feb. 1997.
(8) RGB to YUV [4] P. Hao and Q. Shi, “Comparative study of color transforms for image

0 299 0 587 0 114

: :
00 148 00 289 0 437
: :
0 615 00 515 00 1
: :
:
coding and derivation of integer reversible color transform,” in Proc.
Int. Conf. Pattern Recognition, Sep. 2000, vol. 3, pp. 224–227.
: (53) [5] Y. Chen and P. Hao, “Integer reversible transformation to make JPEG
:
lossless,” in Proc. Int. Conf. Signal Processing, 2004, pp. 835–838.

(9) RGB to RC GC BC [6] P. Hao and Q. Shi, “Matrix factorizations for reversible Integer map-
ping,” IEEE Trans. Signal Process., vol. 49, no. 10, pp. 2314–2324,
1 609 00 447 00 104
: : : Oct. 2001.
00 058 0 977 0 051
: : :
[7] S. Oraintara, Y. J. Chen, and T. Q. Nguyen, “Integer fast Fourier trans-

00 025 00 037 1 162

form,” IEEE Trans. Signal Process., vol. 50, no. 3, pp. 607–618, Mar.
: : : 2002.
(10) RGB to RS GSBS [8] M. D. Adams, F. Kossentini, and R. K. Ward, “Generalized S trans-
1 167 00 146 00 151
: : :
form,” IEEE Trans. Signal Process., vol. 50, no. 11, pp. 2831–2842,

0 114 0 753 0 159

: : : : (54)
Nov. 2002.
[9] K. R. Nagarajan, M. P. Devasahayam, and T. Soundarajan, “Product of
00 001 0 059 1 128
: : : three triangular matrices,” Lin. Algebra Appl., vol. 292, pp. 61–71, Jul.
1999.
For example, from Table I, the process of the forward integer RGB [10] T. Toffoli, “Almost every unit matrix is a ULU,” Lin. Algebra Appl.,
to KLA transform is vol. 259, pp. 31–38, Jul. 1997.
[11] G. Strang, “Every unit matrix is a LULU,” Lin. Algebra Appl., vol. 265,
Step 1)
f [1] = 0 4 [2] + 13 1[1] 16 [2] = 2 [1]
x x = ; f x
pp. 165–172, Nov. 1997.
[12] P. Dita, “Factorization of unitary matrices,” J. Phys. A: Math. Gen.,
f [3] = 0 [3] x : (55) vol. 39, pp. 2781–2789, 2003.
IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 16, NO. 6, JUNE 2007
[13] B. Chen and A. Kaufman, “3D volume rotation using shear trans-
forms,” Graph. Models, vol. 62, pp. 308–322, 2000.
[14] P. Hao, “Customizable triangular factorizations of matrices,” Lin. Al-
gebra Appl., vol. 382, pp. 135–154, May 2004.
[15] S. C. Pei and J. J. Ding, “Reversible integer color transform with bit-
constraint,” in Proc. Int. Conf. Image Processing, 2005, vol. 3, pp.
964–967.
Objective Distortion Measure for Binary Text Image
Based on Edge Line Segment Similarity
Jun Cheng and Alex C. Kot, Fellow, IEEE
Abstract—This paper proposes a new approach to measure the distortion
introduced by changing individual edge pixels in binary text images. The
approach considers not only how many pixels are changed but also where
the pixels are changed and how the flipping affects the overall shape formed
by the edge line. Similarities between the edge line segments in the original
and distorted image are compared to measure the distortion. Subjective
testing shows that the new distortion measure correlates well with human
visual perception.
Index Terms—Binary text image, distortion measure.
I. INTRODUCTION
Distortion measure is an important topic in image processing
including data hiding. An important application of the distortion
measure in data hiding is to evaluate the performance of the data
hiding algorithms as well as to provide insights in data hiding. Dis-
tortion measure can be categorized into subjective measurement and
objective measurement [1]. As discussed in [2], subjective distortion
measure quantifies the dissatisfaction of the viewer in observing the
distorted image in place of the original. A common way to evaluate
the dissatisfaction is through subjective testing. In these tests, observer
views a series of distorted images and rate them based on the visibility
of the artifact. The results of subjective testing depend on various
factors such as the observer’s background. Subjective measurement is
important for image quality evaluation since the images are ultimately
viewed by human beings. However, subjective testing is inconvenient,
time consuming, and expensive. The objective distortion measure gives
the distortion between the original and the distorted image mathemat-
ically such as mean-square error (MSE), peak signal-to-noise ratio
(PSNR), and signal-to-noise ratio (SNR). However, it may not reflect
the observer’s visual perception of distortion. An objective distortion
measure that can accurately reflect the subjective ratings would be
quite useful in designing data hiding algorithms. Several authors have
discussed the gaps between subjective measurement and objective
measurement and they have proposed solutions for multilevel images
Manuscript received April 13, 2006; revised February 4, 2007. The associate
editor coordinating the review of this manuscript and approving it for publica-
tion was Dr. Zhigang (Zeke) Fan.
J. Cheng is with the Workstation Resource Laboratory, School of Electrical
and Electronic Engineering, Nanyang Technological University, Singapore
639798 (e-mail: cjun@pmail.ntu.edu.sg).
A. C. Kot is with the School of Electrical and Electronic Engineering,
Nanyang Technological University, Singapore 639798 (e-mail: eackot@ntu.
edu.sg).
Digital Object Identifier 10.1109/TIP.2007.896619
1057-7149/$25.00 © 2007 IEEE
1691
Fig. 1. Two 3 2 3 pixel patterns with same DRDM distortion score [11].
Fig. 2. Illustration of edges: (a) enlarged “a” and (b) edge line of “a.”
[2]–[4]. Little work has been done for objective measurement for
binary text images. In recently years, data hiding in binary text images
have received much attention and a series of data hiding schemes for
binary images by flipping individually selected edge pixels have been
proposed [5]–[10]. These schemes are called pixel flipping techniques.
One of the important issues of the pixel flipping technique is how to
select the flippable pixels to minimize distortion. Also, the comparison
of the distortion introduced by different schemes is important. How-
ever, performance evaluation is difficult due to the lack of objective
distortion measure that reflects the subjective test results. There is an
urgent need to develop a good objective distortion measure.
Distance reciprocal distortion measure (DRDM) proposed in [11]
uses a weight matrix to calculate the distortion caused by the flipping of
pixels. It shows a better correlation with human perception than PSNR,
and this measure works well if salt and pepper noise is involved. For ex-
ample, flipping a nonedge pixel usually causes a larger visual distortion
than flipping an edge pixel. However, most good data hiding techniques
for binary images involve flipping edge pixels only. If connectivity is
also taken into consideration, the measure of distortion would be more
accurate. For example, flipping the center pixels of the two patterns in
Fig. 1 have the same distortion score according to DRDM, but the visual
distortions by flipping them are quite different. The change in smooth-
ness and connectivity measure (CSCM) by using a modified weight
matrix in [12] gives reasonable distortion scores. However, it has a lim-
itation for patterns such as sharp corner and pixels along straight line.
In this paper, we propose a new objective distortion measure based on
edge line segment to evaluate the distortion introduced by flipping indi-
vidually selected edge pixels for binary text images. The new proposed
method is introduced in Section II, followed by experimental results
and a conclusion.
II. PROPOSED METHOD
We define an edge line as the common sharing ‘line’ between two
neighboring pixels where the pixel values for the two pixels are dif-
ferent. The edge pixel refers to pixels (either white or black) on the
edge. For example, the edge lines in Fig. 2(a) are those black lines
shown in Fig. 2(b). When the edge line changes its direction by 90
degrees, it forms a sharp “corner.” There are eight different types of
corners shown in Fig. 3 that make up all possible corners in a 2 2 2
pixel block.