
TM1300 Data Book
Philips Semiconductors
7-12
PRODUCT SPECIFICATION
7.13.3
YUV-2x Upscaling
In the YUV-2
× modes, the EVO performs 2× horizontal
upscaling of the YUV data from SDRAM. No vertical up-
scaling is performed. The width of the result image
(IMAGE_WIDTH) should be an even number. Upscaling
is performed by 4-tap filtering. For all 3 memory formats,
Y luminance data is upscaled using a (–3,19,19,–3)/32
filter to generate the missing output pixels. Output pixels
at the same location as the input pixels use the corre-
sponding input pixel values, as shown in Figure 7-26.
The U and V chrominance values are generated in the
same way as the Y luminance signal for 2
× upscaling,
assuming that both the input and output use YUV 4:2:2
co-sited chrominance coding. The U and V output pixels
at the same location as the U and V input pixels use the
corresponding input pixel values. The U and V output
pixels between the U and V input pixels are generated
using the (–3,19,19,–3)/32 filter, as shown in Figure 7-26.
If the input chroma is interspersed, a (–1,13,5,–1)/16 fil-
ter is used to generate the U and V output pixels that are
displaced by half a Y pixel from the U and V input pixels,
and a (–1,5,13,–1)/16 filter is used to generate the addi-
tional upscaled U and V output pixels that are displaced
by 1.5 pixels from the U and V input pixels. This is shown
in Figure 7-27.
7.13.4
Pixel Mirroring for Four-tap Filters
The EVO uses a 4-tap filter for upscaling and for convert-
ing from interspersed to co-sited format. One extra pixel
is needed at the beginning and two at the end of each
line processed by this filter. These pixels are supplied
automatically by mirroring the first and last pixels of each
line. For example:
Output pixel 1 uses input pixel 1 to generate its value.
(same location, no ltering).
Output pixel 2 uses pixels 1,1, 2 and 3 to generate its
value.
Output pixel 3 uses pixel 2 to generate its value.
Output pixel 4 pixel uses pixels 1, 2, 3 and 4, etc.
...
Output pixel 2N–2 uses pixels N–2, N–1, N, and N–1
to generate its value.
Output pixel 2N–1 uses pixel N to generate its value.
Output pixel 2N uses pixels N–1, N, N, and N–1 to
generate its value.
Figure 7-28 shows an example of six pixels upscaled to
12 pixels.
WIDTH pixels
HEIGHT
lines
pix0
pix1
pix2
pix
W–1
Y_BASE_ADR
WIDTH/2 pixels
HEIGHT
lines
pix0
pix2
U_BASE_ADR
(Repeated for
V_BASE_ADDR,
V_OFFSET)
Y_OFFSET
U_OFFSET
Figure 7-22. Image storage in planar memory format
for YUV 4:2:2.
Chrominance (U,V)
samples
Luminance
samples
Input Pixels: YUV
Output Pixels: Y’U’V’
Output Location Same
As Input Pixel: Y’U’V’ = YUV
Upscaled Luminance Output Between
Input Pixels: Y’ = (-3,19,19,-3)/32
×Y
Upscaled Chrominance Output Between
Input Pixels: U’,V’ = (-3,19,19,-3)/32
× U,V
Figure 7-26. 2x upscaling of Y pixels.
Chrominance (U,V)
samples
Luminance
samples
Input Pixels: YUV
Output Pixels: Y’U’V’
Co-sited Chrominance Output
U’,V’ = (–1,13,5,–1)/16
×U,V
Co-sited Chrominance Output
U’,V’ = (–1,5,13,–1)/16
×U,V
Upscaled Luminance Output Same
As Input Pixel: Y’ = Y
Upscaled Luminance Output Between
Input Pixels: Y’ = (-3,19,19,-3)/32
× Y
Figure 7-27. 2x upscaling of U and V with interspersed to co-sited conversion.