Samsung S5PC100, User Manual

Page: 1469 / 1863

S5PC100 USER’S MANUAL (REV1.0) MIXER
9.9-17
Table 9.9-1 Graphic Blending-factor Alpha in Case of Normal Mode
MIXER_GRAPHICx_CFG.
REG_WINx_BLEND_EN
MIXER_GRAPHICx_CFGx.
REG_PIXELx_BLEND_EN
ALPHA VALUE
(Blending factor of each pixel)
0 0 1
0 1 blending_factor_each_pixel(A)
1 0
blending_factor_layer
(MIXER_GRAPHICn_CFG[7:0])
1 1
blending_factor_layer *
blending_factor_each_pixel
Table 9.9-2 Graphic Blending Method
Pixel
Blend
Window
Blend
Normal mode Pre multiplied mode
0 0 - -
0 1 Alpha_gw * graphic (1 – alpha_gw) *
lower layer
Alpha_gw * graphic (1 – alpha_gw) *
lower layer
1 0 Alpha_gp * graphic (1 – alpha_gp) *
lower layer
graphic (1 – alpha_gp) * lower layer
1 1 (Alpha_gp * alpha_gw) * graphic (1
– alpha_gp * alpha_gw) * lower layer
alpha_gw * graphic (1 – alpha_gp *
alpha_gw) * lower layer
In pre-multiplied mode, the input graphic data is multiplied by the pixel blending factor (alpha_gp) and
truncated to the size of source format bits. For example, although the result of the multiplication of 8 bit
data by 8 bit pixel blending factor is 16 bits which is the first term of the blending equation in the
Table 9.9-2 Normal mode, the supplied data is truncated to 8 bits that results the loss of the lower significant 8bits
during calculation. In direct 32bpp modes, this loss data is somewhat indistinguishable in visually (+/- 1difference).
But in direct 16bpp modes, the loss data can make visually different result from the original. For example, in 4633
direct mode, the pre-multiplied Cb data is truncated to 3 bits that result in the difference value from +15 to -15.
This difference range can result a visual difference. We can’t reduce the error that is resulted from the 3 bits input
source. So we recommend using the 32bpp mode to use the pre-multiplied mode.