How to get the pixelformat or bitdepth of TPNGImage - delphi

Question

With Delphi 2010 you can get the pixelformat of a jpg file with
TJPEGImage ( Image.Picture.Graphic ).PixelFormat
Is there a way to get the pixelformat or bitdepth of TPNGImage?
I tried this but it returns incorrect bitdepth:
if Lowercase ( ExtractFileExt ( FPath ) ) = '.png' then
StatusBar1.Panels [ 4 ].Text := ' Color Depth: ' + IntToStr( TPNGImage ( Image.Picture.Graphic ).Header.ColorType ) + '-bit';

Answer 1

you must use the BitDepth field
TPNGImage(Image.Picture.Graphic ).Header.BitDepth)
and using the ColorType field you can wirte a function like this
function BitsForPixel(const AColorType, ABitDepth: Byte): Integer;
begin
case AColorType of
COLOR_GRAYSCALEALPHA: Result := (ABitDepth * 2);
COLOR_RGB: Result := (ABitDepth * 3);
COLOR_RGBALPHA: Result := (ABitDepth * 4);
COLOR_GRAYSCALE, COLOR_PALETTE: Result := ABitDepth;
else
Result := 0;
end;
end;
and use like so
procedure TForm72.Button1Click(Sender: TObject);
begin
ShowMessage(IntToStr( BitsForPixel(
TPNGImage ( Image1.Picture.Graphic ).Header.ColorType,
TPNGImage ( Image1.Picture.Graphic ).Header.BitDepth
)));
end;

Answer 2

As Rodrigo pointed out, Header.BitDepth is the value to use. The pitfall is that you have to interpret it depending on the ColorType. You may find some hints in the comments inside function BytesForPixels in PngImage.pas:
{Calculates number of bytes for the number of pixels using the}
{color mode in the paramenter}
function BytesForPixels(const Pixels: Integer; const ColorType,
BitDepth: Byte): Integer;
begin
case ColorType of
{Palette and grayscale contains a single value, for palette}
{an value of size 2^bitdepth pointing to the palette index}
{and grayscale the value from 0 to 2^bitdepth with color intesity}
COLOR_GRAYSCALE, COLOR_PALETTE:
Result := (Pixels * BitDepth + 7) div 8;
{RGB contains 3 values R, G, B with size 2^bitdepth each}
COLOR_RGB:
Result := (Pixels * BitDepth * 3) div 8;
{Contains one value followed by alpha value booth size 2^bitdepth}
COLOR_GRAYSCALEALPHA:
Result := (Pixels * BitDepth * 2) div 8;
{Contains four values size 2^bitdepth, Red, Green, Blue and alpha}
COLOR_RGBALPHA:
Result := (Pixels * BitDepth * 4) div 8;
else
Result := 0;
end {case ColorType}
end;
As you see, for ARGB (= COLOR_RGBALPHA) the BitDepth value is taken for each color part plus the alpha value individually. So BitDepth = 8 will result in a 32-bit value for each pixel.