(PDF) WHT-based composite motion compensated NTSC interframe direct coding - DOKUMEN.TIPS (2024)

(PDF) WHT-based composite motion compensated NTSC interframe direct coding - DOKUMEN.TIPS (1)

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 44. NO. 12. DECEMBER 1996 1711

WHT-Based Composite NTSC Interframe

Motion Compensated Direct Coding

Takahiro Hamada and Shuichi Matsumoto

Abstruct- The motion compensated interframe differential pulse code modulation (DPCM) and discrete cosine transform (DCT) hybrid (MC DCT) coding was nominated as a standard scheme for component TV signals by IS0 and ITU-R. However, in cases where an NTSC composite TV signal is used such as the United States and Japan, applying the MC DCT scheme with its luminancekhrominance separating and composing process causes unavoidable quality degradation. The reason for this additional process required for MC DCT is that a composite TV signal presents a “color subcarrier phase shift problem” in which the color subcarrier phase varies between a coding block and reference block according to the motion vector. In this paper, we propose a Walsh Hadamard Transform (WHT)-based composite motion compensated NTSC interframe direct coding scheme. In this scheme, phase shifts of a color subcarrier and modulated chrominance components between a coding block and reference block can be effectively compensated by a simple process of coefficient permutation and polarity changes of several pairs of WHT coefficients to which 100% of the subcarrier energy and most of the modulated chrominance component’s energy are packed. In the motion compensated DCT scheme, however, the energy of the color subcarrier and modulated chrominance components are spread over too many coefficients and a pair-based coefficient handling rule is not given to solve this problem. This paper demonstrates that the proposed scheme provides higher coding performance for a composite NTSC signal than does the motion compensated DCT scheme with its luminancekhrominance separating and composing process.

I. INTRODUCTION

N THE video coding field, the differential pulse code I modulation (DPCM) and discrete cosine transform (DCT) hybrid coding scheme is currently used as the video coding standard by ISO/ITU because of its high coding efficiency. This scheme requires that the input video signal must be a TV signal with separate luminance and chrominance components. However, the predominant TV signal format used in broadcast studios is still a composite TV signal such as the NTSC used in Japan and the United States, and will remain in use for quite some time. This composite TV signal must, therefore, be sep- arated into luminance and chrominance so component coding can be applied to TV to be broadcast. Component coding, however, has a drawback in that quality degradation occurs in the luminance/chrominance separating and composing process. Picture quality grows even worse in the iterated separating and composing process performed on tandem connections of video codecs for international TV transmission or field pick

Paper approved by B G Haskell, the Editor for SpeecWImage of the IEEE Communications Society Manuscript received March 3, 1995, revised May 24, 1996

The authors are with KDD R&D Laboratories, Saitama, 3.56 Japan Publisher Item Identifier S 0090-6778(96)09026-5

up (FPU). These factors severely reduce the effectiveness of component coding on a composite TV signal. In view of this situation, we conducted intensive research into a direct com- posite coding scheme for a composite TV signal that would not require a luminancekhrominance separating and composing process. Just as with component coding, research into direct composite coding has a long history and much significant research has been reported on DPCM coding [ 11-[5], transform coding [6]-[SI, and comparisons of these coding schemes [9]. With the predominant use of motion-compensated interframe coding [lo]-[ 151, however, the drawback of a “color subcarrier phase shift problem” has appeared in composite direct coding, in which the phase of the color subcarrier phase differs between predictive and coding pixels according to the imotion vector. Motion compensation (MC) trials of a composite TV signal have been conducted under limited conditions with chrominance signals [16], [ 171. However, little research has been done on composite direct coding due to the adoption of the motion-compensated interframe DPCM and DCT hybrid (MC DCT) coding scheme as the standard for component TV signals [18], [19]. The consequence is that broadcast TV signal coding is badly compromised by the current MC DCT scheme, i.e., due to its luminance/chroininance separating process.

In this paper, we propose a Walsh Hadamard Transform (WHT)-based composite motion compensated NTSC inter- frame direct coding. This WHT-based scheme allows direct MC of a composite TV signal with the same efficiency obtained in component coding of component TV signads. We then show that this WHT method delivers higher coding performance in an NTSC signal, than does the MC DCT scheme with its luminance/chrominance separating process. In our WHT direct coding scheme, phase shifts of the color subcarrier and modulated chrominance components between the coding block and a reference block can be compensated for by a simple process of coefficient permutations and polarity changes of several pairs of WHT coefficients. This process is difficult to accomplish with the DCT scheme because the energy of the color subcarrier and the energy of the modulated chrominance components are spread over too many coefficients [20].

The principle of WHT-based composite motion is described in Section 11. A new NTSC interframe direct coding scheme based on this principle is proposed in Section 111. Next, Section IV compares our WHT-based scheme against the motioin com- pensated DCT scheme with its luminance/chrominance sepa- rating process, by means of coding performance analysis using test data in experiments performed by computer simulation.

0090-6778/96$0.5.00 0 1996 IEEE

(PDF) WHT-based composite motion compensated NTSC interframe direct coding - DOKUMEN.TIPS (2)

1712 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 44, NO. 12, DECEMBER 1996

11. COMPOSITE MOTION COMPENSATION

A. Chrominance Signal Expression in WHT

First, the chrominance signal expression in WHT is derived. Then, a field-based 8 pel x 8 line size is taken for the transform block size. This same type of analysis can be conducted for frame-based blocks. Generally, the chrominance signal C of a composite TV signal is found by

EB - EY sin (wsct ) (1) 2.03 cos ( w d ) + ER - EY 1.14

C =

where w,, is an angular frequency of a color subcarrier(CSC) [21]. Equation (1) can be reduced as follows:

C = A COS (w,,t + $I) (2)

with

and EB - EY E R - E Y '

tan$ = -0.5615

A frequency four times that of the CSC frequency ( f s c ) is generally used as the NTSC sampling frequency . In this case, the CSC phase shift between adjacent samples is 90". Let the chrominance signal C(z,g) in a sample of the two-dimensional (2-D) NTSC field ( i : horizontal, j : vertical, i = 1,. . . ,768, j = 1, . . . ,248) with an offset degree 8 be expressed as

(3) I- ' . Y(%J) (7 = 270")

where x ( i , j ) = A ( i , j ) c o s ( + ( i , j ) + 0) and y = A ( i , j ) sin (b( i , j ) + e ) .

To implement an 8 pel x 8 line block-based WHT, if we let a block position (I, J ) be defined as z = 81 + 1, . . . ,8/ + 8 (1 = 0 , . . . ,951, andg = & + I , . . . ,8J+8 ( J = 0 , . . . ,30) , then, the matrix S I , J of an I, J positioned block can be expressed as (4), shown at the bottom of the page.

Next, let 1 x 8 column vectors X I ( J ) and Y I ( J ) be defined by

X I ( J ) = [2 (81+ 1 , ~ ) - 2 ( 8 1 + 2 , j ) 2 ( 8 1 + 3 , J ) -2(81+ 4,~) 2(81+ 5 , ~ ) -2(81+ 6 , ~ ) 2(81+ 7 , j ) -2(81+ 8 , j ) lT

Y I ( J ) =[~(81+ 1 , ~ ) - ~ ( 8 1 + 2 , J ) Y ( 8 1 + 3 , J )

-y(81+ 4,~) y(81+ 5 , ~ ) - ~ ( 8 1 + 6 , ~ )

y ( 8 1 + 7 ) -y(81+ 8,3)IT. ( 5 )

By substituting ( 5 ) for (4), SI ,J can be simply expressed as follows:

S I , J = [ x 1 ( 8 J + 1) Y1(8J + 2) -X1(8J + 3 ) - Y 1 ( 8 5 + 4 ) X1(85 + 5 ) y1(8J + 6 ) - X 1 ( 8 J + 7 ) - Y 1 ( 8 5 t 8 ) ] . (6)

B. Phase Compensation According to Motion Vector

For a motion vector with integer pel accuracy, CSC phase shifts of a reference block S'(Q) from a coding block SI ,J are classified into four types, i.e., Q = 0", 90", 180", and 270", according to a motion vector. Let the motion vector ( M V z , M V y ) be M V x : + in a right-hand direction and MVy: + in the downward direction described in Fig. 1. The relation between CSC phase shifts and ( M V x , M V y ) can then be expressed by

( 4n + 2 (0" shift: @)

(7) 4n (180" shift: 0)

MVz + 2Mvy = { 4n + 1 (270' shift: 0) 4n + 3 (90" shift: 0)

where n is an integer.

SI , J =

(PDF) WHT-based composite motion compensated NTSC interframe direct coding - DOKUMEN.TIPS (3)

HAMADA AND MATSUMOTO: WHT-BASED COMPOSITE MOTION COMPENSATED NTSC INTERFRAME DIRECT CODING 1713

0 GI 0 0 0

Reference oddleven field

to a limited number of coefficients around the DC coefficient. Let it be supposed that in WHT, the horizontal frequency bandwidth of ER-EY and EB-EY exists only among sequencies hl-h4 (Fig. 2). Under this supposition, ER-EY and E~B-EY in a horizontal direction can be expressed by a combination of

G)4@ 0 0 0 0

Current odd/even field

Motion vector MVx,MVy)

f l v- I

H=

I ++++++++

++++---- ++----++ ++--++-- +--++--+ +--+-++-

a# 0, b#O, c#O, d#O : Chrominance components

Fig. 2. Hadamard transform and sequencies with chrominance components.

Fig. 1. Color subcarrier phase shift corresponding to a motion vector. From (9), S'(q) can be rewritten as

From Fig. 1, it is apparent that the CSC phase of a reference block is coincident with that of the coding block in the case M V z + 2MVy = 4n + 2, because CSC changes its polarity per frame.

Due to this fact, if we have a reference block S'(Q) which gives exactly the same chrominance signal (A(zj) and ~ ( Z J ) )

as a coding block SI, J , then S'(Q) becomes

S'(0") = [X1(8J + 1) Y1(8J + 2) -X1(8J + 3 ) -Y1(8J + 4) X1(SJ + 5 ) Y1(8J + 6) -X1(SJ+ 7 ) - Y 1 ( 8 J + 8 ) ]

S'(90") = [ -Y1(8J + 1) X1(8J + 2) Y 1 ( 8 J + 3 ) -X1(8J + 4) - Y 1 ( 8 J + 5 ) X1(8J + 6) Y1(8J+ 7 ) - X 1 ( 8 J + 8 ) ]

S ' ( l 8 O O ) = [-X1(8J + 1) -Y1(8J + 2) X1(8J + 3 ) Y1(8J+4) - X 1 ( 8 J + 5 ) -Y1(8J+ 6)

X1(8J + 7 ) Y 1 ( 8 J + S ) ] S'(270") = [Y1(8J + 1) -X1(8J + 2) -Y,(SJ + 3 )

X1(8J+4) Y1(8J+5) - X 1 ( 8 J + 6 )

- Y r ( S J + 7 ) Xr (8J + S ) ] (8)

by shifting the CSC phase by Oo, 90°, 180°, and 270". Generally, the bandwidth of chrominance signals ER-EY and

EB-EY of a NTSC composite signal is limited to less than about 1.3 MHz. With ER-EY and EB-EY transformed by WHT therefore, the signal power of these components is compacted

S'(0") = [X1(8J + 1) Y 1 ( 8 J + 1) -X1(8J + 3) -Y1(8J + 3 ) X1(8J + 5) Y1(8J + 5 ) -X1(8J + 7 ) - Y r ( 8 J + 7 ) ]

S'(90") = [-Y1(8J + 1) X1(8J + 1) YI(8.J + 3) -X1(SJ + 3 ) -Y1(8J + 5) XI( 85 + 5) Y r ( 8 J + 7 ) -X1(8J+ 7 ) ]

S'(l80") = [-X1(8J + I) -Y1(8J + 1) X 1 ( 8 J + 3 ) Y1(8J+ 3 ) - X 1 ( 8 J + 5 ) - Y 1 ( 8 J + 5 ) X1(8J + 7 ) Y 1 ( 8 J + 7 ) ]

S'(270") = [Y1(8J + 1) -X1(8J + 1) -Y1(8J -t 3)

X1(8J + 3 ) Y 1 ( 8 J + 5 ) -X1(8J + 5 )

- Y 1 ( 8 J + 7 ) X1(8J + 7 ) ] . (10) As the first step, we perform horizontal WHT by multiplying an 8 x 8 WHT matrix H to the right side of S'. The results of four types of S' per column vector are generally shown as

&X1(8J + 1) fY1(8J + 1) fXr(85 + 3 ) kY1(8J + 3 ) f X r ( 8 J + 5 ) fY1(8J + 5 )

X1(8J + 7 ) &Y,r(8J + 7 ) . (1 1) The eight polarities of eight terms of X and Y diiffer per column and these are shown in Table I. In Table I, by the permutation and polarity change of column vectors on S'(90°)H, S'(180°)H, S'(270")H, we can compensate CSC and chrominance phase shifts and make these matrices equal to S'(0")H = S H . This relation between S H and S'H holds for H S H and HS'H after performing vertical WHT. The concept

(PDF) WHT-based composite motion compensated NTSC interframe direct coding - DOKUMEN.TIPS (4)

1714 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL 44, NO 12, DECEMBER 1996

TABLE I COLUMN VECTOR COMPONENTS OBTAINED

BY HORIZONTAL WHT IMPLEMENTATION

of the phase shift compensation according to a motion vector is described in Fig. 3. The column vectors of H S H and HS'H are expressed by 2, and Z;, respectively, in Fig. 3.

In motion compensated DCT coding of component TV signals, prediction efficiency can be improved by MC with fractional pel accuracy. In this scheme also, MC with fractional pixel accuracy can be performed in the following manner. For instance, for the motion vector (e.g. ( M V e , MVy) = (2.5,4.5)), which requires MC with fractional pel accuracy, phase shift compensa- tion is performed on four WHT reference blocks which have integer pixel motion vectors ( ( M V e , M V y ) = (2.0,4.0), (3.0,4.0), (2.0,5.0), (3.0,5.O)).An interpolation among WHT coefficients with the same order as these four reference blocks is then performed. This interpolation in the WHT domain is equivalent to an interpolation in a pixel domain of component coding, and the weighting factor for an interpolation is the same as that in a pixel domain.

Fig. 3 is a general solution of a phase compensation for all WHT coefficients derived only for chrominance compo- nents. This phase compensation may however show reduced performance when the target coefficient has much higher luminance power than chrominance signals. Therefore, before the performance analysis is conducted in Section IV, target coefficients of the phase compensation will be decided using test data.

111. NTSC ~NTERFRAME DIRECT CODING SCHEME

Our proposed NTSC interframe direct coding scheme based on composite MC is shown in Fig. 4. First, WHT is imple- mented for an input NTSC signal and interframe difference values are calculated in WHT. The local decoded image after processing by a quantizerlimverse quantizer is then subjected

........ ......__._._.______ ~ .............

8 X 8 WHT codme block

8

0' Shift @

Fig. 3 . CSC and chrominance signals phase compensation corresponding to a motion vector.

n i WHT

, k;,~:) I Y/c separation Y/C separation Y

Fig 4 Block diagram of NTSC interframe direct coding scheme using WHT WHT denotes Hadamard transform, Q denotes quantization, Qpl denotes inverse quantization, MC denotes motion compensation, ME denotes motion estimation, FM denotes frame memory

to inverse WHT and store in a frame memory with a pixel domain format. In motion estimation such as component MC

(PDF) WHT-based composite motion compensated NTSC interframe direct coding - DOKUMEN.TIPS (5)

HAMADA AND MATSUMOTO: WHT-BASED COMPOSITE MOTION COMPENSATED NTSC INTERFRAME DIRECT CODING 1715

Prediction block

Interpolation + i"- PC

From frame memory

(Reference pixels)

PC: Phase Compensation

Fig. 5 . Block diagram of composite MC.

DCT coding, the motion vector ( M V x , MVy) is detected by block matching with luminance signals between a reference frame in a frame memory and an original frame.

The reference block (pixel) corresponding to the motion vector thus obtained is extracted from the frame memory and fed into the composite MC process. A block diagram of this composite MC process is shown in Fig. 5.

In this diagram, an 8 x 8 WHT is implemented on reference pixels brought from the frame memory to make a prediction block. Phase compensation (PC) of CSC and chrominance signals is then carried out corresponding to the motion vector. In the case of MC with fractional pixel accuracy, a prediction block is produced by performing an interpolation among WHT coefficients of the same order.

Iv . PERFORMANCE ANALYSIS BY COMPUTER SIMULATION

A. Target CoefJicient Determination for Phase Compensation Before conducting a performance analysis of the proposed

scheme, target pairs of coefficients for a phase compensation were determined where chrominance signal (including CSC) is dominant against a luminance signal. We measured this chrominance to luminance power ratio by the method de- scribed in Fig. 6 . First, an input NTSC signal was separated into a luminance signal and a chrominance signal with a 2-D comb filter [22] being used for this Y / C separation. Next, WHT was performed on Y and C, respectively, and the average power of Y(k ,m) and C(k ,m) ( ( k ,m) is a coefficient position) over the whole field was measured for each coefficient, and the ratio (U( k , m)) was calculated.

In Table 11, U ( k , m) (dB) is shown for several types of test data. In this table, the chrominance component becomes more dominant if U ( k , m) has a higher value. In the data for each test, U(8,4) and U ( 8 , 5 ) have extraordinarily high values, and the CSC packs perfectly into the two coefficients. Even though the value for U ( k , m ) varies depending on the test data, the two pairs of coefficients (shaded portions in the table) were a phase compensation target largely satisfying the condition U ( k , m) > 0.

B. Experimental Set Up for t'eqormance Analysis

In this section, a coding performance analysis using NTSC test data is performed, comparing our proposed direct coding scheme (Figs. 4 and 5) with the component MC DCT coding scheme.

An experimental set up for a computer simulation is shown in Fig. 7. NTSC test data was generated by color encoding of ITU-R 4:2:2 standard component TV test data which is widely used for video coding simulations. Data was sampled at 4fsc = 14.32 MHz and our proposed composite MC WHT scheme was compared with the component MC DCT scheme by using test data. A quantization process was performed on these two schemes, and their coding performance then compared by measuring noise power (qL, g i ) and entropy ( e M , e s ) . Linear quantizers with a step size that was uniform regardless of the WHT and DCT coefficient positions were used in quantization process. Next, Y (4fsc sampling) and R- Y/B-Y (2fsc sampling) converted from NTSC test daita were used for the DCT coding input signals. Reference [22] was used in the Y / C separation process. An irrational calculation is also used in the composite-component conversion and a rounding process inserted to provide an integer input and output for DCT coding with 8 b accuracy.

In motion estimation, an area defined by of f 1 5 horizontal pixels and f7 vertical field lines is fully searched with half pixel and half (frame) line accuracy. The vector thus obtained was then provided for both our proposed direct coding scheme (Figs. 4 and 5) and the component MC DCT scheme.

Entropy of the composite/component TV signal is defined as follows. Let p(r , k , m) be the probability of the representative quantization value r (k , m) of a WHT coefficient ]position (k ,m) and the composite TV signal entropy e M is then

8 8 8

k=l m = l r=-R

For the component TV signal entropy e s , let p(r1, k , m)/p ( r2 , k , m)/p ( r3 , k , m) be the probability of the representative quantization value rl(r1 = -R1, . . . , 0,

- R3, . . . , 0, . . . , R3). Entropy is calculated for each component Y/R-Y/B-Y according to its sampling rate, and added up because each component is usually coded independently. For, e, we then obtain

. . . , R l ) / r 2 ( ~ 2 = -R2,. . . , 0 , . . . , R 2 ) / ~ 3 ( ~ 3 =

8 8 f R 1

The signal-to-noise ratio (SNR) is calculated by taldng the difference between the original and decoded image obtained

(PDF) WHT-based composite motion compensated NTSC interframe direct coding - DOKUMEN.TIPS (6)

1716 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 44, NO 12, DECEMBER 1996

TABLE I1 LUMINANCEKHROMINANCE POWER RATIO (dB) ON FIELD-BASED WHT DOMAIN

-103

-84

-76

Cheer leaders I1 Boats I1 Autumn leaves

-85 -66 -38 -25 -51 -61 -55 -120 -91 -72 -47 -32 -60 -70 -63 -109 -94 -70 -40 -17 -52 -69 -61

-74 -55 -26 -17 -43 -5.5 -43 -93 -19 -62 -34 -23 -49 -63 -52 -91 -83 -61 -29 -10 -45 -62 -49

-70 -51 -21 -15 -41 -51 -40 -83 -73 -53 -26 -21 -45 -59 -47 -83 -77 -59 -29 -12 -42 -58 -46

7 . t 5

-64 r 6 2 . .- . ...

-60 -60

-49 -21 -14 -41 -50 -38 -79 -67 -49 -22 -17 -43 -56 -44 -79 -72 -54 -25 -10 -40 -55 -43

-41 -8 -1 -30 -47 -32 -71 -63 -43 -14 -5 -37 -53 -40 -66 -64 -40 -6 8 -27 -47 -30

-38 -6 1 -28 -46 -30 -69 -60 -41 -11 -1 -32 -50 -37 -63 -63 -36 -2 9 -25 -44 -26

-116 -87 -70 -39 -39 -58 -67 -60

-88 -80 -60 -30 -20 -49 -63 -52

-82 , -77 , -51 , -28 , -28 , -46 , -60 , -49

after a quantization and inverse quantization process, as

-112 -90 -70 -40 -30 -56 -68 -61 -110 -84 -70 -43 -26 -58 -68 -58

-92 -82 -65 -35 -24 -50 -62 -54 -87 -78 -61 -32 -18 -46 -63 -48

, -20 , -47 , -60 , -48 ,, -78 , -73 , -54 , -25 , -15 , -43 , -60 , -46 , , -83 , -17 , -60 , -31

2552 SNR 1 1010g,O - q2

-79

-72

-68

with

qh (composite MC WHT) q 2 = { q i (component MC DCT)

where q; includes distortion due to component-composite conversion.

I

-13 -55 -17 -21 -44 -59 -47 -78 -72 -56 -21 -17 -44 -57 -46 -74 -70 -50 -21 -15 -43 -58 -45

-70 -48 -10 -17 -38 -56 -44 -69 -68 -47 -13 -6 -35 -52 -39 -66 -66 -45 -32 -1 -33 -54 -35

-67 -44 -10 -3 -34 -53 -39 -68 -65 -44 -12 -3 -33 -49 -36 -63 -63 -41 -9 3 -29 -SO -31

C. Experimental Results and Discussion Noise (difference) power (q&, q i ) and entropy ( e M , e,)

are calculated and shown in Fig. 8 as SNR versus bit-rate, following Fig. 7 by varying the quantizer step size from 1 to 65 at Q for all transform coefficients. As test data, “mobile and calendar” and “flower garden” were selected. In Fig. 8, the

composite MC WHT (proposed method) gives a higher coding performance than component MC DCT in both test data, and this superiority in performance also differs according to bit- rate. Namely, about a 2-3 dB improvement is obtained at a low bit-rate of 0.5-1 .O b/pel and more than 5.0 dB at a high bit-rate of 2.0 b/pel. The reason for this is as follows: in Fig. 8, an upper limiting SNR for component MC DCT coding is drawn as a dotted line. This limit is generated by a double rounding up process of composite-component conversion. The component coding performance is saturated toward this upper limiting SNR as bit-rate increases. Composite coding, however, is free from the rounding process and does not exhibit saturated performance. Noise due to the rounding process is, therefore, more dominant than that of the quantization process at high bit-rate and the difference in performance widens steadily between the two curves. Conversely, quantization noise is more dominant at low bit-rate, but improvements of 3 dB in

(PDF) WHT-based composite motion compensated NTSC interframe direct coding - DOKUMEN.TIPS (7)

HAMADA AND MATSUMOTO: WHT-BASED COMPOSITE MOTION COMPENSATED NTSC INTERFRAME DIRECT CODING 1717

r separation

2D comb filter

Average

Average

Fig. 6. Measuring of Ir/C relative power ratio.

SNR vs. bit-rate compari!

eM1 A " 0

NTSC I test data Composite

Mvx ,Mvy)

Component

Composite

Fig. 7. The concept of computer simulation experiment.

mobile and calendar, and 2 dB in flower garden were still obtained at even less than 1.0 b/pel. This can be explained by the fact that in a composite TV signal, both luminance and chrominance signals are expressed on a common WHT block

and the correlation between these components can be utilized for bit reduction while it is impossible for component MC DCT in which these components are handled by completely separate DCT blocks. The advantage provided by NTSC direct coding

(PDF) WHT-based composite motion compensated NTSC interframe direct coding - DOKUMEN.TIPS (8)

1718 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 44, NO. 12, DECEMBER 1996

45

40 CompositeMCWHT -

25

-

0.0 1 .o 2.0 3.0 4.0 Bit-rate (bit/pei)

(Flower garden)

50 I Upper bound SNR for component cording

Composite MC WHT

20 L L - - 0.0 1.0 2.0 3.0 4.0 5.0 6.0

Bit-rate (bit/pel) (Mobile & calendar)

Fig. 8. Coding performance comparison.

becomes more obvious in the abundant test data available in saturated color signals such as mobile and calendar.

We can conclude from these results that composite MC WHT is more suited for application to NTSC coding than is component MC DCT with its luminance/chrominance sep- aration.

V. CONCLUSION

We have proposed a WHT-based motion compensated interframe direct coding scheme for a NTSC composite signal which is the current dominant format for a TV signal. In our scheme, phase shifts of color subcarrier and chrominance signals are efficiently compensated so that they correspond to the motion vector. Results of computer simulation show that a coding gain of 2-3 dB can be obtained at a low bit-rate (0.5-1.0 b/pel) and more than 5 dB can be obtained at high bit-rate (higher than 2.0 b/pel).

We can conclude from these results that even though MC DCT coding is known to be suitable for a component TV signal format such as HDTV signals, the WHT-based coding scheme proposed here is superior for a composite TV signal format mainly used in SNG and FPU.

ACKNOWLEDGMENT

The authors would like to thank H. Murakami and K. Koga for their valuable suggestions and encouragements, and the anonymous reviewers for their useful comments.

REFERENCES

[ 11 J. E. Thompson, “Differential encoding of composite color television signals using chrominance-corrected prediction,” IEEE Trans. Commun., vol. COM-22, pp. 1106-1 113, Aug. 1974.

[2] K. Sawada and H. Kotera, “32Mbitls transmission of NTSC color TV signals by composite DPCM coding,” IEEE Trans. Commun., vol. COM-26, pp. 1432-1439, Oct. 1978.

[3] Y. Hatori and H. Yamamoto, “Predictive coding for NTSC composite color television signals based on comb-filter integration method,” Trans. IECE, vol. E62, pp. 201-208, Apr. 1979.

[4] H. Yamamoto, Y. Hatori, and H. Murakami, “30 Mbit/s codec for the NTSC color TV signal using an interfield-intrafield adaptive prediction,” IEEE Trans. Commun., vol. COM-29, pp. 1859-1867, Dec. 1981.

[5] R. C. Brainard, A. N. Netravali, and D. E. Pearson, “Predictive coding of composite NTSC color television signals,” SMPTE L, vol. 91, pp. 245-252, Mar. 1982.

[6] T. Ohira, M. Hayakawa, K. Matsumoto, and K. Shibata, “Picture quality of Hadamard transform coding using nonlinear quantizing for color television signals,” presented at the Symposium on the Application of Walsh Functions, Washington D.C., Apr. 1973.

[7] H. F. Harmuth, H. C. Andrews, and K. Shihata, “Two-dimensional sequency filters,” IEEE Trans. Commun., vol. COM-20, no. 3, June 1972.

181 C. Ekambaram and S . C. Kwatra, “A new architecture for adaptive transform compression of NTSC composite video signal,” in Proc. NTC’XI, New Orleans, LA, 1981, pp. C9.6.1-9.6.5

[9] H. Murakami, Y. Hatori, and H. Yamamoto, “Comparison between DPCM and Hadamard transform coding in the composite coding of the NTSC color TV signal,” IEEE Trans. Commun., vol COM-30, pp. 469479, Mar. 1982.

[lo] T. Koga, K. Iinuma, A. Hirano, Y. Iijima, and T. Ishiguro, “Motion compensated interframe coding for video conferencing,” in Proc. NTC 81, New Orleans, LA, 1981, pp. G5.3.1LG5.3.5.

[11] S. Brofferio, C. Cafforio, M. Piacentini, and F. Rocca, “Motion com- pensation in teleconference,” in Proc. ICC 82, Philadelphia, PA, June 1982, pp. 2G.6.1-2G6.5.

[ 121 J. R. Jain and A. K. Jain, “Displacement measurement and its application in interframe image coding,” IEEE Trans. Commun., vol. COM-29, pp. 1799-1808, Dec. 1981.

[13] R. Srinvasan and K. R. Rao, “Predictive coding based on efficient motion estimation,” in Proc. ICC’84, Amsterdam,kolland, May 1984, pp. 521-526.

[I41 C. M. Lin and S. C. Kwatra, “Motion compensated interframe color image coding,” in Proc. ICC’84, Amsterdam, Holland, May 1984.

[15] K. A. Prabhu and A. N. Netravali, “Motion compensated component color coding,” IEEE Trans. Commun., vol. COM-30, pp. 2519-2526, Dec. 1982.

[16] K. A. P rabhu and A. N. Netravali, “Motion compensated composite color coding TV signals,” IEEE Trans. Commun., vol. COM-31, pp. 216-223, Feb. 1983.

[ 171 S . Sabri, “Movement compensated interframe prediction for NTSC color TV signals,” IEEE Trans. Commun., vol. COM-32, pp. 954-968, Aug. 1984.

[I81 ISO-IEC/JTCl/SC29/WGlI, “Coded Representaion of Picture and Au- dio Information,” MPEG93/225b.

[I91 CCIR Study Groups, “Draft Revision to Recommendation 723; Transmission of Component Coded Digital Televison Signals for Contribution-Quality Applications at the Third Hierarchical Level of CCITT Recommendation G. 702,” CMTT-Z/TEMP/20-E.

[20] S. Matsumoto, T. Hamada, M. Saito and H. Murakami: “45 Mbps multi- channel composite TV coding system,” IEICE TRANS. Commun., vol. E75-B, no. 6, 358-367, May 1992.

[21] A. N. Netravali and B. G. Haskel, DigitalPictures. New York Plenum, 1988.

[22] J. P. Rossi, “Digital TV comb filter with adaptive features,” in IERE Con$ Video Data Recording, July 1976.

(PDF) WHT-based composite motion compensated NTSC interframe direct coding - DOKUMEN.TIPS (9)

HAMADA AND MATSUMOTO WHT-BASED COMPOSITE MOTION COMPENSATED NTSC INTERFRAME DIRECT CODING 1719

Takahiro Hamada was born in Nara, Japan on July 25, 1961 He received the B.S degree in electrical engineering from Tokyo University, Tokyo, Japan in 1985, and the M.S degree in electrical engineering from the California Institute of Technology, Pasadena in 1989

Since 1987, he has been with the Research and Development Laboratories of KDD (Overseas Telecommunication Co Ltd of Japan), Tokyo, Japan, and has worked on the development of digital coding of broadcast TV and high-definition - - I

TV signals His research interests include transform coding, quantization technologies, motion compensation, and various kinds of filtering techniques for high quality and highly efficient television signal coding. His work in these areas has resulted in the development of a Japanese coding standard for an enhanced wide screen TV (EDTV 11)

Shuichi Matsumoto was born in Hakodate, Japan, in 1954. He received the B.S., M.S., and Dr. Eng. degrees, from Hokkaido Universit) , Sapporo, Japan, all in electronic?, in 1977, 1979, and 1990, respectively.

Since 1979, he has been with Research and Development Laboratories of KDD (Overseas Telecommunication Co. Ltd. of Japan), Tokyo, Japan, and has worked on the development of digital coding for broadcast TV and higki-definition TV signals. Currently, he is a Senior Manager at

Dr. Matsumoto received the 1983 Best Paper Award and the 1986 Sinohara the Visual Communications Laboratory of KDD R&D Labs.

Memorial Award from IEICE Japan.

(PDF) WHT-based composite motion compensated NTSC interframe direct coding - DOKUMEN.TIPS (2024)
Top Articles
Latest Posts
Article information

Author: Neely Ledner

Last Updated:

Views: 5589

Rating: 4.1 / 5 (62 voted)

Reviews: 85% of readers found this page helpful

Author information

Name: Neely Ledner

Birthday: 1998-06-09

Address: 443 Barrows Terrace, New Jodyberg, CO 57462-5329

Phone: +2433516856029

Job: Central Legal Facilitator

Hobby: Backpacking, Jogging, Magic, Driving, Macrame, Embroidery, Foraging

Introduction: My name is Neely Ledner, I am a bright, determined, beautiful, adventurous, adventurous, spotless, calm person who loves writing and wants to share my knowledge and understanding with you.