Why don't getframe RGB codes agree with plotted polyshape RGB codes?

In the code below, I have computed the RGB color code of the square as seen in both a polyshape plot (polyshapeRGB) and as seen by a getframe() capture of this plot (getframeRGB). Why are the two not in agreement, and is there a way either to make them agree or to predict one version of the RGB values from the other?

h=plot(nsidedpoly(4)); 
I=getframe; 

ctr=round(size(I.cdata,1:2)/2);

polyshapeRGB=round(255*h.FaceColor)

polyshapeRGB = 1×3

0 114 189

getframeRGB(1:3)=I.cdata( ctr(1), ctr(2),:)

getframeRGB = 1×3

166 205 231

[Walter Roberson]

h=plot(nsidedpoly(4)); 
I=getframe; 

ctr=round(size(I.cdata,1:2)/2);

polyshapeRGB=round(255*h.FaceColor)

polyshapeRGB = 1×3

0 114 189

getframeRGB(1:3)=I.cdata( ctr(1), ctr(2),:)

getframeRGB = 1×3

166 205 231

figure
polyshapeRGB = reshape(uint8(polyshapeRGB), 1, 1, 3);
image(repmat(polyshapeRGB, 64, 64, 1));
title('polyshape swath')

figure
getframeRGB = reshape(uint8(getframeRGB), 1, 1, 3);
image(repmat(getframeRGB, 64, 64, 1));
title('getframe swath')

Answer 1

[Voss]

They don't agree because of the transparency of the polyshape.

Default FaceAlpha=0.35:

figure
h=plot(nsidedpoly(4)); % default FaceAlpha=0.35
I=getframe();

ctr=round(size(I.cdata,1:2)/2);
polyshapeRGB=round(255*h.FaceColor)

polyshapeRGB = 1x3

0 114 189

getframeRGB(1:3)=I.cdata( ctr(1), ctr(2),:)

getframeRGB = 1x3

166 205 231

FaceAlpha=1, fully opaque:

figure
h=plot(nsidedpoly(4),'FaceAlpha',1); % fully opaque
I=getframe();

ctr=round(size(I.cdata,1:2)/2);
polyshapeRGB=round(255*h.FaceColor)

polyshapeRGB = 1x3

0 114 189

getframeRGB(1:3)=I.cdata( ctr(1), ctr(2),:)

getframeRGB = 1x3

0 114 189

[Matt J]

Yes, I see that now, thanks. But how is the alpha-blending done? Is it always a linear blending?

It seems to be in this case,

alpha=0.35;

predictedRGB = round( [0   114   189]*alpha + 255*(1-alpha) )

predictedRGB = 1×3

166 206 232

[Voss]

Seems linear in that case, yeah. Another case where it seems linear:

figure
h=plot(nsidedpoly(4)); % default FaceAlpha=0.35
alpha = h.FaceAlpha;
polycolor = h.FaceColor*255;
axescolor = [255 120 50];
set(gca(),'Color',axescolor/255);
I=getframe();

ctr=round(size(I.cdata,1:2)/2);
getframeRGB(1:3)=I.cdata( ctr(1), ctr(2),:)

getframeRGB = 1x3

166 117 98

predictedRGB = round( polycolor*alpha + axescolor*(1-alpha) )

predictedRGB = 1x3

166 118 99

So I think always linear would be a reasonable guess.