HORZ
SIZE
IN
RSZ
width
input
]
0
:
12
[
_
_
_
=
HORZ
SIZE
OUT
RSZ
width
output
]
0
:
11
[
_
_
_
=
VERT
SIZE
IN
RSZ
height
input
]
16
:
28
[
_
_
_
=
î
í
ì
=
=
=
)
422
(
0
]
27
[
_
,
2
)
(
1
]
27
[
_
,
1
_
_
YUV
INPTYP
CNT
RSZ
when
separate
color
INPTYP
CNT
RSZ
when
pixel
per
bytes
camisp-E097
Public Version
Camera ISP Basic Programming Model
www.ti.com
•
Larger than 4:1 downsampling: Although it is rarely necessary to generate a very small image from a
large image, this is supported by the hardware. For example, 10x downsampling can be realized first
with 4x downsampling on-the-fly with the preview engine, then 2.5x downsampling in the memory-input
path. There may not be much time outside the active data region for the second pass, but since the
image is already reduced to 1/16 of its original size, not much time is necessary. Typically, sensor or
video input has 10 ~ 20 % of usable vertical blanking.
For all these scenarios, the second pass can be configured and initiated from an interrupt service routine
triggered by the resizer end-of-frame interrupt:
[24] RSZ_DONE_IRQ (or
6.5.8.5.2 Camera ISP Resizer Processing Time Calculation
The time calculated below is the time it takes for all resizes where the input source is memory (second
pass when doing a 10x resize in preview mode; in the case of a 10x resize in preview mode, the first pass
is hidden behind the time it takes to capture and process the image from the sensor based on cam_pclk
and the number of lines resized).
The following equation can be used to determine the processing time of the resizer when the input is from
memory and therefore how much time it takes before it can switch back to preview input mode:
Time = [ W × bytes_per_pixel × H]/[L3/2]
(4)
Where:
/* If the input is YUV422 and horizontal downsampling is performed: */
if ((
[27] INPTYP ==0) && ((
[9:0] HRSZ + 1) >256))
W = average (input_width, output width); /* output width includes extra 4 pixels if edge enhancement is
enabled*/
else
W = max(input width, output width); /*output width includes extra 4 pixels if edge enhancement is
enabled*/
This time is the baseline steady state calculation of the hardware and does not include the time it takes for
the hardware to fetch the first input and fill the Video processing hardware. It also does not include the
time spent from the last output from the resizer to get back to memory when the resizer interrupt occurs.
However, these beginning and ending times are relatively negligible.
Depending on real-time constraints, this processing time may be much faster than is required. (data
fetches can be delayed from memory to free more bandwidth for use by other system peripherals; see
, Input from Memory.
6.5.8.6
Camera ISP Resizer Summary of Constraints
The following is a list of register configuration constraints to adhere to when programming the resizer. It
can be used as a quick checklist. More detailed register setting constraints can be found in the individual
register descriptions.
•
Vertical and horizontal resize ratio values must be within the following range: [64..1024].
•
Output width:
–
Must be within the maximum limit:
1288
Camera Image Signal Processor
SWPU177N – December 2009 – Revised November 2010
Copyright © 2009–2010, Texas Instruments Incorporated