How to draw a translucent image on a form? - delphi

I want to a draw a translucent image on a Delphi form, but for some reason it is not working.
Here is the original PNG (border is semi transparent):
I load the image in a TImage object:
Image1.Transparent := True;
Form1.Color := clWhite;
Form1.TransparentColor := True;
Form1.TransparentColorValue := clWhite;
The application:
The image isn't translucent. I am working with a BMP image that contains the alpha channel. Am I missing something?

I found a solution that will let you draw a BMP image with an alpha channel onto a form using only the Windows API:
const
AC_SRC_OVER = 0;
AC_SRC_ALPHA = 1;
type
BLENDFUNCTION = packed record
BlendOp,
BlendFlags,
SourceConstantAlpha,
AlphaFormat: byte;
end;
function WinAlphaBlend(hdcDest: HDC; xoriginDest, yoriginDest, wDest, hDest: integer;
hdcSrc: HDC; xoriginSrc, yoriginSrc, wSrc, hSrc: integer; ftn: BLENDFUNCTION): LongBool;
stdcall; external 'Msimg32.dll' name 'AlphaBlend';
procedure TForm4.FormClick(Sender: TObject);
var
hbm: HBITMAP;
bm: BITMAP;
bf: BLENDFUNCTION;
dc: HDC;
begin
hbm := LoadImage(0,
'C:\Users\Andreas Rejbrand\Skrivbord\RatingCtrl.bmp',
IMAGE_BITMAP,
0,
0,
LR_LOADFROMFILE);
if hbm = 0 then
RaiseLastOSError;
try
if GetObject(hbm, sizeof(bm), #bm) = 0 then RaiseLastOSError;
dc := CreateCompatibleDC(0);
if dc = 0 then RaiseLastOSError;
try
if SelectObject(dc, hbm) = 0 then RaiseLastOSError;
bf.BlendOp := AC_SRC_OVER;
bf.BlendFlags := 0;
bf.SourceConstantAlpha := 255;
bf.AlphaFormat := AC_SRC_ALPHA;
if not WinAlphaBlend(Canvas.Handle,
10,
10,
bm.bmWidth,
bm.bmHeight,
dc,
0,
0,
bm.bmWidth,
bm.bmHeight,
bf) then RaiseLastOSError;
finally
DeleteDC(dc);
end;
finally
DeleteObject(hbm);
end;
end;
Using The GIMP, I converted the PNG image
found here to a 32-bit RGBA bitmap, found here, and the result is very good:

Why not try do draw your png onto new image with regular bmp. Draw what you want onto image 2 and redraw /or assign/ all to your image 1 when finish. Must works...

The TransparentColorValue approach cannot possibly work, because this only works with images in which a single colour represents full transparency. [In addition, you are toying with the form's transparent colour instead of image's transparent colour!] The above PNG image is supposed to have an alpha channel, so it's not like every pixel is either shown or transparent -- instead, each pixel has an opacity between 0 and 1 (0.37, for instance). That is, in addition to the R, G, and B components of each pixel, there is an 'alpha' component A.
The above image appears to be corrupt, however. A 'correct' PNG is shown below:
(source: rejbrand.se)
You can try to blend the above one onto different backgrounds, and you will find that the shadow blends nicely.
So, if one has a 'correct' PNG, how to draw it onto a form? Well, that is going to be very difficult in your case, since Delphi 7 does not support PNG images. It only supports BMP images, and these normally do not have alpha channels.

Related

How to use correctly TBitmap object to save a file with transparency?

Below is my sample code:
var lBitmap: TBitmap;
begin
lBitmap := TBitmap.Create;
lBitmap.PixelFormat := TPixelFormat.pf32bit;
lBitmap.Transparent := TRUE; // !
lBitmap.LoadFromFile( 'd:\temp\bmp32b_300dpi_transparent_400x250.bmp' );
// Bitmap RGB+Alpha created with GIMP
// modifications on pixels
Canvas.Draw(100, 0, lBitmap);
// Up to this point it is correct, the drawing is painted with transparency
lBitmap.SaveToFile( 'd:\tmp\after.bmp' ); // after this -> I have lost transparency
lBitmap.Free;
end;
How to use correctly TBitmap object to save a file with transparency?
It seems to me like TBitmap doesn't support saving bitmaps with alpha channels. And maybe we shouldn't blame the VCL for this, because BMPs with alpha transparency are uncommon. Many applications don't support transparent BMPs.
This being said, I "reverse-engineered" a BMP with alpha channel created in GIMP and wrote the following Delphi routine to produce the very same bitmap:
procedure SaveTransparentBitmap(ABitmap: TBitmap; const AFileName: string);
var
FS: TFileStream;
BFH: TBitmapFileHeader;
BIH: TBitmapV5Header;
y: Integer;
sl: PUInt64;
begin
// ABitmap MUST have the GIMP BGRA format.
FS := TFileStream.Create(AFileName, fmOpenWrite);
try
// Bitmap file header
FillChar(BFH, SizeOf(BFH), 0);
BFH.bfType := $4D42; // BM
BFH.bfSize := 4 * ABitmap.Width * ABitmap.Height + SizeOf(BFH) + SizeOf(BIH);
BFH.bfOffBits := SizeOf(BFH) + SizeOf(BIH);
FS.Write(BFH, SizeOf(BFH));
// Bitmap info header
FillChar(BIH, SizeOf(BIH), 0);
BIH.bV5Size := SizeOf(BIH);
BIH.bV5Width := ABitmap.Width;
BIH.bV5Height := ABitmap.Height;
BIH.bV5Planes := 1;
BIH.bV5BitCount := 32;
BIH.bV5Compression := BI_BITFIELDS;
BIH.bV5SizeImage := 4 * ABitmap.Width * ABitmap.Height;
BIH.bV5XPelsPerMeter := 11811;
BIH.bV5YPelsPerMeter := 11811;
BIH.bV5ClrUsed := 0;
BIH.bV5ClrImportant := 0;
BIH.bV5RedMask := $00FF0000;
BIH.bV5GreenMask := $0000FF00;
BIH.bV5BlueMask := $000000FF;
BIH.bV5AlphaMask := $FF000000;
BIH.bV5CSType := $73524742; // BGRs
BIH.bV5Intent := LCS_GM_GRAPHICS;
FS.Write(BIH, SizeOf(BIH));
// Pixels
for y := ABitmap.Height - 1 downto 0 do
begin
sl := ABitmap.ScanLine[y];
FS.Write(sl^, 4 * ABitmap.Width);
end;
finally
FS.Free;
end;
end;
This write a BITMAPFILEHEADER followed by a BITMAPV5HEADER and the pixel data in BGRA format.
I omit all kinds of error checking. For instance, I don't verify that ABitmap actually has the required format.
Test:
procedure TForm1.FormCreate(Sender: TObject);
var
bm: TBitmap;
begin
bm := TBitmap.Create;
try
bm.LoadFromFile('C:\Users\Andreas Rejbrand\Desktop\Test.bmp');
SaveTransparentBitmap(bm, 'C:\Users\Andreas Rejbrand\Desktop\Test2.bmp');
finally
bm.Free;
end;
end;
After this, Test.bmp and Test2.bmp are binary equal.
Saving 32-bit bitmaps with alpha channels needs a workaround as #Andreas Rejbrand has pointed out. There also seems to be some more confusion about the BMP file format, what the TBitmap.Transparent property does, and how you draw bitmaps transparently with the VCL.
32-bit bitmaps are the only bitmaps that carry information about transparency in the files. They have that information in the alpha channel and nowhere else. In the alpha channel, every pixel has its own 0-255 alpha value in the RGBA structure. This is often referred to as partial transparency.
When you draw/display 32-bit bitmaps, you have to pay attention to the TBitmap.AlphaFormat property. It defaults to afIgnore, which means that the bitmap is drawn without transparency. Use afPremultiplied or afDefined to draw with transparency. The latter is probably what you want.
The TBitmap.Transparent property is specific to the VCL TBitmap, and there is nothing in the BMP file format that corresponds to it. It's just a simple way to display bitmaps transparently, where a color defines which pixels should be fully transparent. The application must be familiar with the bitmaps to be able to use this method. It's also important to be aware of how the TBitmap.TransparentMode property works. It defaults to tmAuto, which sets the color of the bottom-leftmost pixel of the bitmap as TBitmap.TransparentColor. When TransparentMode is set to tmFixed, the TBitmap.TransparentColor you have specified is used. This method can also be used on 32-bit bitmaps.
Note that when you draw with the standard VCL TCanvas drawing routines on a 32-bit bitmap with transparency in the alpha channel, the transparency will be lost where you have drawn.
It seems that in your sample code, you have ignored AlphaFormat and TransparentMode. You should also decide if you want to use the transparency in the alpha channel or the TBitmap.Transparent method. But we have no bitmap in order to check if that is the real problem.

Saving transparent (alpha channel) PNG from TImageList

I have a TImageList which contains transparent icons (32bit, with alpha channel). What I want to do is to save individual icons based on image index as PNG file(s), while preserving alpha channel transparency. Using RAD Studio 2010 so it has TPngImage support, no need for third party libraries. Images are loaded into TImageList from PNG "sprite" image using the method here - Add a png image to a imagelist in runtime using Delphi XE - so the transparency is preserved upon loading. Now I need to save them out individually, in other words, extract individual images from sprite images which is already loaded into TImageList.
My code so far:
int imageindex = 123;
boost::scoped_ptr<TPngImage> png(new TPngImage);
boost::scoped_ptr<Graphics::TBitmap> bmp(new Graphics::TBitmap);
MyImageList->GetBitmap(imageindex, bmp.get()); // Using GetBitmap to copy TImageList image into separate TBitmap
png->Assign(bmp.get()); // Assign that bitmap to TPngImage
png->SaveToFile("C:\\filename.png");
The above works but it saves with the white background (transparency is not preserved after saving). I am probably missing a simple step but can't figure it out.
Delphi code is also welcome, shouldn't be hard to translate.
Yes, you can obtain PNG-image from TImageList where it was added. Code below allows you to do this!
Firstly, add PngImage to your uses clause.
procedure LoadPNGFromImageList(AImageList: TCustomImageList; AIndex: Integer; var ADestPNG: TPngImage);
const
PixelsQuad = MaxInt div SizeOf(TRGBQuad) - 1;
type
TRGBAArray = Array [0..PixelsQuad - 1] of TRGBQuad;
PRGBAArray = ^TRGBAArray;
var
ContentBmp: TBitmap;
RowInOut: PRGBAArray;
RowAlpha: PByteArray;
X: Integer;
Y: Integer;
begin
if not Assigned(AImageList) or (AIndex < 0) or
(AIndex > AImageList.Count - 1) or not Assigned(ADestPNG)
then
Exit;
ContentBmp := TBitmap.Create;
try
ContentBmp.SetSize(ADestPNG.Width, ADestPNG.Height);
ContentBmp.PixelFormat := pf32bit;
// Allocate zero alpha-channel
for Y:=0 to ContentBmp.Height - 1 do
begin
RowInOut := ContentBmp.ScanLine[Y];
for X:=0 to ContentBmp.Width - 1 do
RowInOut[X].rgbReserved := 0;
end;
ContentBmp.AlphaFormat := afDefined;
// Copy image
AImageList.Draw(ContentBmp.Canvas, 0, 0, AIndex, true);
// Now ContentBmp has premultiplied alpha value, but it will
// make bitmap too dark after converting it to PNG. Setting
// AlphaFormat property to afIgnored helps to unpremultiply
// alpha value of each pixel in bitmap.
ContentBmp.AlphaFormat := afIgnored;
// Copy graphical data and alpha-channel values
ADestPNG.Assign(ContentBmp);
ADestPNG.CreateAlpha;
for Y:=0 to ContentBmp.Height - 1 do
begin
RowInOut := ContentBmp.ScanLine[Y];
RowAlpha := ADestPNG.AlphaScanline[Y];
for X:=0 to ContentBmp.Width - 1 do
RowAlpha[X] := RowInOut[X].rgbReserved;
end;
finally
ContentBmp.Free;
end;
end;
Look at the picture. It is depicts what will happen if we set or not set such line of code:
ContentBmp.AlphaFormat := afIgnored;
Figure 1 is a result of setting afIgnored and the second one figure is a result of not setting afIgnored, allowing to use previously set afDefined.
Original image is an image named Figure 1
Using of code above in application:
procedure TForm1.aButton1Click(Sender: TObject);
var
DestPNG: TPngImage;
begin
DestPNG := TPNGImage.Create;
try
// Initialize PNG
DestPNG.CreateBlank(COLOR_RGBALPHA, 8, 60, 60);
// Obtain PNG from image list
LoadPNGFromImageList(ImageList1, 0, DestPNG);
// Output PNG onto Canvas
DestPNG.Draw(Canvas, Rect(0, 0, 60, 60));
DestPNG.SaveToFile('C:\MyPNGIcon.png');
finally
DestPNG.Free;
end;
end;

Blending semi-transparent bitmaps containing text with Graphics32

I'm trying to implement a layered painting system inside one of our internal components and I have problems blending bitmaps containing text.
Following code fragment shows the problem:
uses
GR32;
procedure DrawBitmaps;
var
bmp1: TBitmap32;
bmp2: TBitmap32;
begin
bmp1 := TBitmap32.Create;
bmp1.Width := 100;
bmp1.Height := 100;
bmp1.FillRect(0, 0, 100, 100, clWhite32);
bmp1.FillRect(0, 0, 80, 80, clTrGreen32);
bmp1.Font.Size := -16;
bmp1.Font.Color := clBlack;
bmp1.TextOut(2, 10, 'Green');
bmp1.SaveToFile('c:\0\bmp1.bmp');
bmp2 := TBitmap32.Create;
bmp2.Width := 80;
bmp2.Height := 80;
bmp2.FillRect(0, 0, 80, 80, clTrRed32);
bmp2.Font.Size := -16;
bmp2.Font.Color := clBlack;
bmp2.TextOut(2, 50, 'Red');
bmp2.SaveToFile('c:\0\bmp2.bmp');
bmp2.DrawMode := dmBlend;
bmp2.DrawTo(bmp1, 20, 20);
bmp1.SaveToFile('c:\0\bmpcombined.bmp');
bmp1.Free;
bmp2.Free;
end;
Resulting images:
bmp1:
bmp2:
bmpcombined:
As you can see, text is painted in black on bmp and bmp2, but appears white on bmpcombined.
I'm guessing the problem lies in TextOut which maps to Windows.ExtTextOut (via GR32_Backends_VCL.pas, TGDIBackend.Textout). That method doesn't handle transparency and paints text with alpha 00 (color is $000000 instead of $FF000000).
As a quick fix, setting bmp2.Font.Color to $FF000000 doesn't help.
bmp2.Font.Color := TColor(clBlack32);
I am using fresh sources from GitHub
How should I paint a non-transparent text on a semi-transparent background so that I could blend this into a larger picture?
As far as I remember the TextOut function was only meant as a direct way to add some text to the bitmap, lacking all the fixes you mentioned above.
In order to remain full control over transparency you might want to use
procedure TBitmap32.RenderText(X, Y: Integer; const Text: string; AALevel: Integer; Color: TColor32);
instead.
It uses the techniques you mentioned in your own answer, but on a more sophisticated level. It also allows you to use anti-aliasing (based on oversampling), but today it's not actually recommended to use any other anti-aliasing technique other than what the font-engine outputs (to take full advantage of font-hinting).
As you are using the latest source code you could also consider to use VPR to render the text (see example 'TextVPR'). It converts the outline of the text as vectors and use the vector drawing capabilities of Graphics32 (by default the engine 'VPR' is used, hence the name) to render it onto the screen. There's also a stripped down engine of AGG included, which itself is based on the FreeType1 engine, which might be slightly faster for fonts.
Speaking of performance: Keep in mind that every other approach than TextOut will obviously decrease the performance. So if you aim for high performance might better cook your own code (based on TextOut).
Otherwise the TextVPR approach leaves you with more freedom, especially when it comes to filling the text (with a gradient for example) or transforming the text (to a curve or such).
The best solution I could find requires three helper functions.
TransparentToOpaque changes all fully transparent pixels to opaque.
procedure TransparentToOpaque(bmp: TCustomBitmap32);
var
I: Integer;
D: PColor32Entry;
begin
D := PColor32Entry(#bmp.Bits[0]);
for I := 0 to bmp.Width * bmp.Height - 1 do begin
if D.A = 0 then
D.A := $FF;
Inc(D);
end;
bmp.Changed;
end;
FlipTransparency changes all fully transparent pixels to opaque and vice versa.
procedure FlipTransparency(bmp: TCustomBitmap32);
var
I: Integer;
D: PColor32Entry;
begin
D := PColor32Entry(#bmp.Bits[0]);
for I := 0 to bmp.Width * bmp.Height - 1 do begin
if D.A = 0 then
D.A := $FF
else if D.A = $FF then
D.A := 0;
Inc(D);
end;
bmp.Changed;
end;
MakeOpaque marks all pixels as opaque.
procedure MakeOpaque(bmp: TCustomBitmap32);
var
I: Integer;
D: PColor32Entry;
begin
D := PColor32Entry(#bmp.Bits[0]);
for I := 0 to bmp.Width * bmp.Height - 1 do begin
D.A := $FF;
Inc(D);
end;
bmp.Changed;
end;
Following tricks can then be applied.
After drawing text on the main image bmp1 which doesn't contain transparent pixels, code calls TransparentToOpaque to prevent problems with blending later on.
When drawing on a (semi)transparent bitmap bmp2, code creates yet another bitmap bmp3 and fills it with an opaque version of that (semi)transparent bitmap. This will ensure that font is aliased to correct colors in the TextOut call.
After the TextOut bmp3 contains opaque background and transparent text. FlipTransparency is then called to generate opaque text on a transparent background.
bmp3 is blended onto bmp2. This gives up opaque text on a (semi)transparent background.
bmp2 is blended onto bmp1.
Example code:
procedure TForm53.DrawBitmaps;
var
bmp1: TBitmap32;
bmp2: TBitmap32;
bmp3: TBitmap32;
begin
bmp1 := TBitmap32.Create;
bmp1.Width := 100;
bmp1.Height := 100;
bmp1.FillRect(0, 0, 100, 100, clWhite32);
bmp1.FillRect(0, 0, 80, 80, clTrGreen32);
bmp1.Font.Size := -16;
bmp1.Font.Color := clBlack;
bmp1.TextOut(2, 10, 'Green');
//Mark all fully transparent pixels (generated with TextOut) as opaque.
TransparentToOpaque(bmp1);
SaveBitmap32ToPNG(bmp1, 'c:\0\bmp1a.png');
bmp2 := TBitmap32.Create;
bmp2.Width := 80;
bmp2.Height := 80;
bmp2.FillRect(0, 0, 80, 80, clTrRed32);
//Create bitmap, large enough to contain drawn text (same size as original bitmap in this example).
bmp3 := TBitmap32.Create;
bmp3.Width := bmp2.Width;
bmp3.Height := bmp2.Height;
//Copy `bmp2` to `bmp3`.
bmp2.DrawMode := dmOpaque;
bmp2.DrawTo(bmp3, 0, 0);
//Mark all pixels as opaque (alpha = $FF)
MakeOpaque(bmp3);
//Draw text on `bmp3`. This will create proper aliasing.
bmp3.Font.Size := -16;
bmp3.Font.Color := clBlack;
bmp3.TextOut(2, 50, 'Red');
//Make all fully transparent pixels (TextOut) opaque and all fully opaque pixels
// (background coming from `bmp2`) transparent.
FlipTransparency(bmp3);
SaveBitmap32ToPNG(bmp3, 'c:\0\bmp3a.png');
//Blend `bmp3` on semi-transparent background (`bmp2`).
bmp3.DrawMode := dmBlend;
bmp3.DrawTo(bmp2, 0, 0);
SaveBitmap32ToPNG(bmp2, 'c:\0\bmp2a.png');
//Blend background + text onto main image.
bmp2.DrawMode := dmBlend;
bmp2.DrawTo(bmp1, 20, 20);
SaveBitmap32ToPNG(bmp1, 'c:\0\bmpcombineda.png');
bmp1.Free;
bmp2.Free;
bmp3.Free;
end;
Resulting images:
bmp1a:
bmp2a:
bmp3a:
bmpcombineda:

Saving ImgView32 transparent layers to PNG

I have problems saving an ImgView32 layer as a TRANSPARENT PNG.
I use the code from this question to do the saving.
However, the image saves with a white background.
Here is how I initialize my ImgView32, create a layer on it, and then draw a line on it:
procedure TputLine.FormCreate(Sender: TObject);
var
P: TPoint;
W, H: Single;
imwidth: integer;
imheight: integer;
begin
imwidth := Iv1.Width;
imheight := Iv1.Height;
with iv1 do
begin
Selection := nil;
Layers.Clear;
Scale := 1;
Scaled := True;
Bitmap.DrawMode := dmTransparent;
Bitmap.SetSize(imwidth, imheight);
Bitmap.Canvas.Pen.Width := 4;
end;
BL := TBitmapLayer.Create(iv1.Layers);
try
BL.Bitmap.DrawMode := dmTransparent;
BL.Bitmap.SetSize(imwidth,imheight);
BL.Bitmap.Canvas.Pen.Width := penwidth;
BL.Bitmap.Canvas.Pen.Color := pencolor;
BL.Location := GR32.FloatRect(0, 0, imwidth, imheight);
BL.Scaled := False;
except
BL.Free;
raise;
end;
end;
So iv1 is the name of my ImgView32.
Then I draw a line on it using this code:
var
bm32:TBitmapLayer;
...
begin
bm32:=(iv1.Layers[0] as TBitmapLayer).Bitmap;
bm32.canvas.pen.color:=clwhite;
bm32.canvas.brush.color:=clwhite;
bm32.canvas.rectangle(0,0,bm32.width-1, bm32.height-1);
bm32.canvas.Pen.Color:=WinColor(ColorPickerGTK1.SelectedColor);
bm32.canvas.brush.color:=clWhite;
bm32.Canvas.Pen.Width:=3;
bm32.Canvas.MoveTo(0,bm32.Height);
bm32.Canvas.LineTo(0+150,bm32.Height-250);
end;
If I use the clWhite32 for the above code when drawing the rectangle, then when saving the PNG, the background of the imgView turns black... So I do not understand the problem really.
I do the saving like this:
procedure TputLine.Button2Click(Sender: TObject);
var
myLay:TBitmapLayer;
begin
mylay := iv1.Layers.Items[0] as TBitmapLayer;
SavePNGTransparentX(mylay.Bitmap);
end;
and the actual saving code (from the link described above)
procedure TPutLine.SavePNGTransparentX(bm32:TBitmap32);
var
Y: Integer;
X: Integer;
Png: TPortableNetworkGraphic32;
function IsWhite(Color32: TColor32): Boolean;
begin
Result:= (TColor32Entry(Color32).B = 255) and
(TColor32Entry(Color32).G = 255) and
(TColor32Entry(Color32).R = 255);
end;
begin
bm32.ResetAlpha;
for Y := 0 to bm32.Height-1 do
for X := 0 to bm32.Width-1 do
begin
if IsWhite(bm32.Pixel[X, Y]) then
bm32.Pixel[X,Y]:=Color32(255,255,255,0);
end;
Png:= TPortableNetworkGraphic32.Create;
Png.Assign(bm32);
Png.SaveToFile('C:\ThisShouldBeTransparent.png');
Png.Free;
end;
I do not understand why it does not save the layer as transparent PNG.
How can I fix it? Any idea is welcome.
You can replicate my problem using the above code. It uses GR32_PNG and GR32_PortableNetworkGraphic. You only need to add a TImgView32 control to your form and add the code listed above.
The reason to the problem seems to be two-fold.
First, when you call Png.Assign(bm32); in unit GR32_PNG, it attempts to find out what the smallest format would be to store the image in. If the image has less than 256 distinct colors, it creates a paletted format, and depending on how many colors it finds, the bit depth can become 1, 2, 4 or 8. As far as my knowledge goes, only images with TrueColor and Alpha can be saved as variable transparency png images.
Secondly, you draw with only one color, which triggers the above problem. This is of course not your fault, IMO the above mentioned analysis should be possible to bypass.
The TPortableNetworkGraphic32 class has two properties, BitDepth and ColorType wich control the format of the png image, that would be useful, if they were settable! Attempting to set them in code as:
Png.BitDepth := 8;
Png.ColorType := ctTrueColorAlpha;
leads to exceptions
EPngError with message 'Bit depth may not be specified directly yet!
EPngError with message 'Color Type may not be specified directly yet!
From the wording we can assume some further development in the future.
The cure
To bypass the above described image analysis, you can change line 459 in GR32_PNG.pas.
procedure TPortableNetworkGraphic32.AssignPropertiesFromBitmap32()
var
...
begin
...
IsPalette := True; // <--- change to False
...
That will take care of the bit depth analysis and prevents palette creation if less than 256 colors.
After this hack you can use the SavePNGTransparentX() procedure to save a TBitmap32 to a .png file and preserving transparency.
Then, there's one change more that you may be interested in, regarding the SavePNGTransparentX() procedure. As you have seen, it requires the background of your drawing surface to be white, because it specifically sets the Alpha channel to zero for all white pixels. A TBitmapLayer is however initialized with all pixels (RGBA) as all zeros, so the color components of each pixel makes up to black color (which is not visible because the alpha channel is zero). Therefore you need to fill the drawing layer with white, which makes it opaque, which again covers up all lower layers (beneath the drawing layer).
To make a correction to this you can
remove the initialization of the drawing layer to all white
change the IsWhite function into an IsBlack function
change the assignment of the transparent pixels
Code would become
procedure TForm8.SavePNGTransparentX(bm32:TBitmap32);
var
Y: Integer;
X: Integer;
Png: TPortableNetworkGraphic32;
function IsBlack(Color32: TColor32): Boolean;
begin
Result:= (TColor32Entry(Color32).B = 0) and
(TColor32Entry(Color32).G = 0) and
(TColor32Entry(Color32).R = 0);
end;
function IsWhite(Color32: TColor32): Boolean;
begin
Result:= (TColor32Entry(Color32).B = 255) and
(TColor32Entry(Color32).G = 255) and
(TColor32Entry(Color32).R = 255);
end;
begin
bm32.ResetAlpha;
for Y := 0 to bm32.Height-1 do
for X := 0 to bm32.Width-1 do
begin
// if IsWhite(bm32.Pixel[X, Y]) then
// bm32.Pixel[X,Y]:=Color32(255,255,255, 0);
if IsBlack(bm32.Pixel[X, Y]) then
bm32.Pixel[X,Y]:=Color32( 0, 0, 0, 0);
end;
Png:= TPortableNetworkGraphic32.Create;
Png.Assign(bm32);
Png.SaveToFile('C:\tmp\imgs\ThisShouldBeTransparent3.png');
Png.Free;
end;
With this change you can see the layers beneath your drawing layer and the resulting file of your drawing layer will have all non-drawn pixels transparent.
There is however still one problem in general with the above procedure, and that is that partially transparent pixels loose their transparency, but that will remain for a future excercise.
Here's a few images, First the ImageView with a bitmap loaded on the bottom layer:
Then, I draw a blue cross on the BL layer (ref. your code) using basically the code in your Button1Click() but without the white rectangle.
Then I save the BL layer to a .png file and look at it with Windows Explorer "Preview":

TPNGObject - Create a new blank image and draw translucent images on it

I am building an application that has "virtual windows". The output is TImage object.
1) The application loads window skin files into TPNGObject's:
2) Then application has to create a new blank TPNGObject, and resize the skin files to needed sizes and draw them on that blank image. Should look something like this:
3) And the final output on TImage:
The problem is that I do know how to create a completely blank off screen image. Of course I could simply render the skin files on to TImage each time, but it's easier and better to resize skin files and create the window once, instead.
I'm using the PNG Library by Gustavo Daud, version 1.564 (31st July, 2006).
The below uses CreatePNG procedure of 'pngfunctions.pas' of Martijn Sally, from an extension library (pngcomponents) to pngimage.
var
Bmp, Mask: TBitmap;
PNG: TPNGObject;
begin
Bmp := TBitmap.Create;
Bmp.PixelFormat := pf24bit;
Bmp.SetSize(64, 64);
Bmp.Canvas.Brush.Color := clBtnFace;
Bmp.Canvas.Font.Color := clRed;
Bmp.Canvas.Font.Size := 24;
Bmp.Canvas.TextOut(4, 10, 'text');
Mask := TBitmap.Create;
Mask.PixelFormat := pf24bit;
Mask.Canvas.Brush.Color := clBlack;
Mask.SetSize(64, 64);
Mask.Canvas.Font.Color := clWhite;
Mask.Canvas.Font.Size := 24;
Mask.Canvas.TextOut(4, 10, 'text');
PNG := TPNGObject.Create;
CreatePNG(Bmp, Mask, PNG, False);
PNG.Draw(Canvas, Rect(10, 10, 74, 74));
// finally, free etc...
Here's the output (black, white squares are TShapes):
My other answer is another alternative which I suggest. However your question still poses an issue: The PNG library must either have a bug which is preventing any canvas drawing from being visible (after using CreateBlank constructor with COLOR_RGBALPHA as color type) or we're all missing something.
It looks like the only workaround that I can see is (as you mention in your edit) use a Bitmap to do your drawing instead. Use the transparent properties of this bitmap (Transparent: Bool and TransparentColor: TColor) to set up the transparent area of your image, then when you need a transparent PNG, just copy that bitmap over to the new PNG object...
BMP.Width:= 100;
BMP.Height:= 100;
BMP.Transparent:= True;
BMP.TransparentColor:= clWhite;
BMP.Canvas.Brush.Style:= bsSolid;
BMP.Canvas.Brush.Color:= clWhite;
BMP.Canvas.FillRect(BMP.Canvas.ClipRect);
BMP.Canvas.Brush.Color:= clBlue;
BMP.Canvas.Ellipse(10, 10, 90, 90);
PNG.Assign(BMP);
And the white area of the image should be transparent. There are other ways of accomplishing the transparent area, but that's another subject.
Image:
Is this what you're trying to do?
I apologize to people that I messed their heads up.
It turns out CreateBlank works as wanted. The problem was that I was drawing PNG on PNG canvas (PNG.Canvas.Draw). Canvas doesn't really support transparency. To draw a translucent PNG on another PNG you will need a procedure/function that merges those both layers together. With some googling I ended up with this procedure:
procedure MergePNGLayer(Layer1, Layer2: TPNGObject; Const aLeft, aTop: Integer);
var
x, y: Integer;
SL1, SL2, SLBlended: pRGBLine;
aSL1, aSL2, aSLBlended: PByteArray;
blendCoeff: single;
blendedPNG, Lay2buff: TPNGObject;
begin
blendedPNG := TPNGObject.Create;
blendedPNG.Assign(Layer1);
Lay2buff:=TPNGObject.Create;
Lay2buff.Assign(Layer2);
SetPNGCanvasSize(Layer2, Layer1.Width, Layer1.Height, aLeft, aTop);
for y := 0 to Layer1.Height - 1 do
begin
SL1 := Layer1.Scanline[y];
SL2 := Layer2.Scanline[y];
aSL1 := Layer1.AlphaScanline[y];
aSL2 := Layer2.AlphaScanline[y];
SLBlended := blendedPNG.Scanline[y];
aSLBlended := blendedPNG.AlphaScanline[y];
for x := 0 to Layer1.Width - 1 do
begin
blendCoeff:=aSL1[x] * 100/255/100;
aSLBlended[x] := round(aSL2[x] + (aSL1[x]-aSL2[x]) * blendCoeff);
SLBlended[x].rgbtRed := round(SL2[x].rgbtRed + (SL1[x].rgbtRed-SL2[x].rgbtRed) * blendCoeff);
SLBlended[x].rgbtGreen := round(SL2[x].rgbtGreen + (SL1[x].rgbtGreen-SL2[x].rgbtGreen) * blendCoeff);
SLBlended[x].rgbtBlue := round(SL2[x].rgbtBlue + (SL1[x].rgbtBlue-SL2[x].rgbtBlue) * blendCoeff);
end;
end;
Layer1.Assign(blendedPNG);
Layer2.Assign(Lay2buff);
blendedPNG.Free;
Lay2buff.Free;
end;
Usage:
var
PNG1, PNG2: TPNGObject;
begin
PNG1 := TPNGObject.CreateBlank(COLOR_RGBALPHA, 16, 500, 500);
PNG2 := TPNGObject.Create;
PNG2.LoadFromFile('...*.png');
MergePNGLayer(PNG1, PNG2, 0, 0);
// PNG1 is the output
And again, I am really sorry to users that wanted to help, but couldn't due to not understanding me.
I don't have answer to this question but I figured out how to get same result with any PNG editor. I have created blank 1000x1000 PNG image and saved it in my application directory. Then I open this image in my program and resize image to needed sizes (smaller of course) and that's the trick.

Resources