Renesas SH7781 User Manual

Page 1050

Advertising
background image

20. Graphics Data Translation Accelerator (GDTA)

Rev.1.00 Jan. 10, 2008 Page 1020 of 1658
REJ09B0261-0100

(1)

Estimated Image Generation Function

The following shows an outline of the estimated image generation function.

Output frame

Y pointer (current)

Output frame

U pointer (current)

Output frame

V pointer (current)

Past frame

Y pointer

Past frame

U pointer

Past frame

V pointer

Future frame

Y pointer

Future frame

U pointer

Future frame

V pointer

(9) Estimated
image writing

(5) Data
reading

Image data for one frame

Frame width

Y padding

U (current)

V (current)

U (past)
V (past)

Y (future)

U (future)
V (future)

Y (past)

Y (current)

Base point

Base point

Output

(current)

Output

target position

Target position (mbcol, mbrow)

Comparison point

(+Recon_right, down)

(6) Half-pixel correction processing

Recon_down

(ne

g

ative value indicates reverse direction)

Invalid data

Y: 16

× 16 dot

U: 8

× 8 dot

V: 8

× 8 dot

Input
image
(half-pixel)

Recon_right (negative value indicates reverse direction)

Y: 17

× 17 dot

U: 9

× 9 dot

V: 9

× 9 dot

For half-pixel processing,
get data for a block with one
extra dot according to motion
vector values

Hardware performs processing
as is even when in padding part

Y0 (128 bytes)
Y1 (128 bytes)
Y2 (128 bytes)
Y3 (128 bytes)

U (128 bytes)
V (128 bytes)

Buffer RAM 1

(8) Estimated-image generation

(7) IDCT data
reading

IDCT data

Y: 16

× 16 dot

U: 8

× 8 dot

V: 8

× 8 dot

Y: 16

× 16 dot

U: 8

× 8 dot

V: 8

× 8 dot

Y: 17

× 17 dot

U: 9

× 9 dot

V: 9

× 9 dot

Y: 16

× 16 dot

U: 8

× 8 dot

V: 8

× 8 dot

Correction processing result (Y: 16

× 16, U/V: 8 × 8)

Data after half-pixel

correction

IDCT data

Data to be written back to the current target position
(result of forward macroblock processing)

Ч 16

Ч 16

Ч 16

Ч 16

Ч 16

Ч 16

Ч 17

Ч 17

Input data for one macroblock

Correction
case 3

Correction case 4

Correction

case 1

Correction

case 2

Each frame pointer should point to an address on a 16-byte boundary
for Y and 8-byte boundary for U and V.

8-bit saturation
calculation
(0

≤ x ≤ 255)

(1) Calculation of output position (first row)

(2) Calculation of output position (nth row)

(3) Calculation of input position (first row)

(4) Calculation of input position (nth row)

A

B

E

D

A'

B'

D'

C'

a

b

d

c

A''

B'

D'

C'

mbcol

mbrow

Frame hei

g

ht

(mbcol, mbrow)

DDR2-SDRAM

Figure 20.6 Outline of Estimated Image Generation Function

Advertising