AN2621 Freescale Semiconductor / Motorola, AN2621 Datasheet - Page 3

• Manufacturer Part Number: AN2621
• Description: MPC8220i PF300 Image Coprocessor Operation
• Manufacturer: Freescale Semiconductor / Motorola

1.2
The RU functions by stepping through the destination pixel array and, with each step, incrementing a pair
of high-precision registers that contain the coordinates of the destination pixel transformed back to the
coordinate system of the source image. Each time the destination pixel location is incremented by one pixel,
the location of the new source pixel is calculated by adding the contents of two registers to the contents of
the X dimension source register and the contents of two other registers to the contents of the Y dimension
source register. The values of the four summing registers represent the unit vector in the destination image
transformed back to the source image space. The X and Y dimension source registers are initialized with the
coordinates of the upper left corner of the destination image (1,1) transformed back into the source image
coordinate system. Thirty two bits each of integer and fraction (64 bits in total) represent the X and Y
dimension source image coordinate registers and each of the four registers that are used to increment them
to assure sufficient range and precision. The initial contents of Xsrc and Ysrc and the values of the four
summing registers can be found by inverting the system of equation 1.
Once the required source pixel coordinates are found, they are translated into the linear address space of the
input pixel array so that the pixel values can be fetched from source memory and loaded into the next stage
of the pipeline, the Color Conversion Unit (CCU). The PF300 asserts itself as memory bus master in order
to read the required data from SDRAM. Any resulting source image coordinate that falls outside of the
source image is marked with a 0 in the “shadow” image mask indicating an invalid pixel value in the output
buffer.
A data flow chart of the RU is shown in Appendix A, “Resampling Unit Flow Chart,” and a C language
source code listing of the RU implementation in the MPC8220i is supplied in Appendix B, “Resampling
Unit C Code.”
2
2.1
The Color Conversion Unit (CCU) performs a lookup and piecewise interpolation in a sparse
three-dimensional lookup table. A sparse lookup table is one in which only a subset of the possible inputs
are represented by data. Outputs for intermediate values are calculated by interpolation between the points
that are supplied. The CCU computes the interpolation in dedicated hardware.
Even though the table can be addressed with as many as 16,777,215 (2
individual dimension only has 256 possible addresses. The CCU takes advantage of this fact by
precomputing many intermediate results and pointers that can then be accessed for use in real time by simple
lookups into several much smaller (256 element) tables. The use of this block allows the three-dimensional
lookup with a table that is much smaller than the data size used to address it. Up to four tables can be loaded
and interpolated simultaneously, each addressed by the same data triplet. This allows the interpolation of all
components for four colorant printing systems with a single pass of the color data through the pipeline.
Examples of color space conversions that can be performed using this method include the following:
MOTOROLA
•
•
•
•
•
•
CIE Lab −> CMYK
CIE Lab −> RGB
RGB −> CMYK
RGB −> CIE Lab
RGB −> K
CMYK −> K
Computational Details
Color Conversion Unit (CCU)
Algorithm Description
MPC8220i PF300 Image Coprocessor Operation
Freescale Semiconductor, Inc.
For More Information On This Product,
Go to: www.freescale.com
24
) different color combinations, each
Color Conversion Unit (CCU)
3