pnx1700 NXP Semiconductors, pnx1700 Datasheet - Page 366

• Manufacturer Part Number: pnx1700
• Description: Connected Media Processor
• Manufacturer: NXP Semiconductors

Philips Semiconductors
Volume 1 of 1
Table 3: Chroma Key ROP Examples
PNX17XX_SER_1
Preliminary data sheet
ROP
0xFF
0x00
0x80
0xF0
0xE8
Description
All ROP bits are enabled, so all of the key combinations are included. One of them must be true; therefore, the
result will always be true (chroma key=1).
None of the ROP bits is enabled, so the ROP will never get a match. The result will always be false (chroma
key=0).
Bit 7 is enabled. This is the combination where all three keys=1. The result will be true only when all the pixel’s
YUV (RGB) are in the YUV (RGB) chroma key range.
Bits 7,6,5 and 4 are enabled. These are the combinations where Key2=1. For any pixel where the Y(R)
component (Key2) matches the chroma key range, the ROP result will be 1.
Bits 7, 6, 5, and 3 are enabled (11101000=E8). These are the combinations where at least two of the keys are
true. In this case, the ROP will return true whenever any two of the color components match their respective
chroma key range.
The ROP should be programmed to enable each key combination for which one
wants to trigger a “color key=true.” For each pixel in the stream, the ROP checks the
keys to determine if they match one of the selected combinations. When there is a
match, the result is true (1). If there is no match, the result is false (0). The result of
the Color Key Combining ROPs is combined in the mixer and used as a key (Key1) in
the Invert, Pixel Select, Alpha Blend Select and Pass Through Key ROPs.
The examples in
The result of the four ROPs within a layer is fed into the mixer associated with the
layer. The four keys are called: Current_Color_Key_0-3.
UDTH
UDTH is the undither unit that is used to recover 9 bits of precision from an 8-bit
dithered data. It is the reverse of the dither operation in the VIP as an 8-bit-wide video
is usually not very practical (since some head room is lost because there is not a
good automatic gain control). For quality reasons, therefore, one has to process 9 or
10-bit video.
However, if every stage in the processing chain—after its required processing—has
to round to 8 bits, accumulated quantization errors (quantization noise) occurs. The
local video data paths can be made wider (e.g., 10 bits), but since there are only 8-bit
wide field memories, compression from 9 bits or more to 8 bits or less will have to
take place. Furthermore, if pixels are blindly rounded and/or quantized to 8 bits,
particularly for low-frequency (small) signals, the quantization noise is not evenly
distributed but remains correlated to the input signal, and contouring effects occur.
So, what is wanted is 9-bit video quality for an 8-bit (memory) price. With this goal in
mind, it is primarily to de-correlate the quantization error and also to retain some
precision, that dithering is used. Dithering distributes the error across the entire
spectrum simply by adding some random noise prior to quantization via a random or
semi-random perturbation of the pixel values. The 8-bit values, with 9-bit precision,
are stored in memory where they are fetched from and processed by the QVCP. For
the part of a picture that is almost constant (or flat) in the horizontal direction, one
should try to recover the full 9 bits from the stored 8 bits because the quantization
error is more noticeable; However, when there are more high-frequency components,
the full 9 bits cannot be recovered, but the quantization error is less visible anyway.
Table 3
Rev. 1 — 17 March 2006
describe how to program the ROP for various results.
© Koninklijke Philips Electronics N.V. 2006. All rights reserved.
PNX17xx Series
Chapter 11: QVCP
11-9