LF2301JC25 Ver la hoja de datos (PDF) - LOGIC Devices

Número de pieza

componentes Descripción

Fabricante

LF2301JC25

Image resampling sequencer. Speed 55ns

LOGIC Devices 

DEVICES INCORPORATED

LF2301

Image Resampling Sequencer

Image Rotation & Bilinear Interpolation

Figure 8 shows an example of rotating

an image 30º and using bilinear

interpolation. This mode is selected by

loading M1-0 with “00” for a clockwise

spiral walk. A counterclockwise spiral

walk could be selected by loading M1-0

with “10.” Bilinear interpolation

requires a kernel size of 2 x 2 pixels.

Loading K3-0 with “0001” selects a

kernel size of 2 x 2. The first pixel

selected is determined by x0 and y0. In

this example, the first pixel is (0,0). In

this case, the LF2301s should address

consecutive pixels during each spiral

walk. For this to occur, FOV must be

set to 1 (F2-0 loaded with “001”).

After the last pixel of a spiral walk has

been selected, the next pixel address is

determined by adding dx/du to the

current X address and by adding

dy/du to the current Y address

(unless the kernel just completed was

the last for that line). At the end of the

first spiral walk, pixel (0,1) is ad-

dressed. Since the next calculated

pixel should be (0.866,0.5), dx/du is

selected to be 0.866 and dy/du is

selected to be 0.5. However, after

adding dx/du and dy/du to the X

and Y addresses respectively, the

generated address is (0.866,1.5). The Y

address is off by a value of 1. This is

due to the fact that the last pixel

address of a spiral walk is used to

calculate the first pixel address of

the next spiral walk. In order for the

LF2301s to generate the correct result,

dy/du must be modified by subtract-

ing a 1 from it. The correct value of

dy/du is -0.5. Figure 6 shows how the

unmodified differential terms were

calculated.

After the last pixel of the last spiral

walk on the first line has been selected,

the first pixel address of the second

line is determined by adding dx/dv to

x0 and by adding dy/dv to y0. Since

the first calculated pixel of the first

spiral walk on the second line should

be (-0.5,0.866), dx/dv is selected to

be -0.5 and dy/dv is selected to be

0.866. Second order differential terms

are not used in this transform and are

therefore set to 0.

It is important to note that the integer

portion of the address generated in

the LF2301 is used as the X or Y

pixel address. The fractional portion

(sub-pixel portion) is used as the

coefficient RAM address.

UMIN and VMIN are both selected to

be 0. UMAX and VMAX are both

selected to be 2. Table 7 shows the

values loaded into all Parameter

Registers. Table 8 shows the ITS

outputs for this example.

FIGURE 6. DIFFERENTIAL TERMS

dx = cos 30º = 0.866

du

dy

du

= sin 30º

= 0.5

dx

dv

= –sin 30º= –0.5

dy

dv

= cos 30º = 0.866

FIGURE 7. 30° IMAGE ROTATION

dx

dx

dv

du

dy

dv

length

30º

=1

v

y

source image pixels

calculated pixels

x

dy

du

u

FIGURE 8. 30° IMAGE ROTATION

-1

0

1

2

3

5

1,2

0

3

4

1

2

3

4

1 = 1st pixel of 1st spiral walk,

2 = 1st pixel of 2nd spiral walk, etc.

source image pixels

calculated pixels

TABLE 7. PARAMETER REGISTERS

ADDR Row (HEX) Column (HEX)

0000

000

000

0001

FFF

FFF

0010

000

000

0011

000

100

0100

DDB

800

0101

100

1FF

0110

800

0111

1FF

DDB

100

1000

000

000

1001

000

000

1010

000

000

1011

000

000

1100

000

000

1101

000

000

1110

000

000

1111

002

002

Video Imaging Products

2-9

08/16/2000–LDS.2301-H

Share Link: