Alpha Channel

Color Keys

Color key is a color which is interpreted as transparent when displaying or blending. This technique is widely used in DirectX programming, however, Graphics Mill for .NET does not use it (it uses alpha blending instead). It is used only when Graphics Mill for .NET saves indexed image into a format which supports transparency only in color key mode (such as GIF).

The main disadvantage of this method is that it cannot represent semitransparency. That's why colorkeying cannot produce smooth blend on the edge of opaque and transparent area. Another problem of this approach is that if you appoint some color to be transparent, you cannot display this color on this image. It will be replaced by transparent pixels.

Due to these disadvantages Graphics Mill for .NET uses another method: alpha blending.

Alpha Channel

Instead of making some color transparent, each pixel stores information on how opaque is it. This opacity data is called alpha channel and is typically stored after color channels of the pixel. The most wide-spread alpha-compliant pixel format is Format32bppArgb. Each channel (including alpha channel) occupies one byte. Alpha channel is stored in the high (fourth) byte.

When alpha = 255 (max value for this pixel format), the color is treated as completely opaque, when alpha = 0, the color is completely transparent (i.e. when blending this bitmap with another one, background color is not changed). For alpha values between 0 and 255, color is interpreted as semi-transparent and during blending each channel of the resulting color is calculated using the following formula:

result = (foreground*alpha + background*(max-alpha))/max

where max is 255 for 8 bit per channel pixel format, and 65355 for 16 bits per channel (extended) pixel formats.

Working with Alpha Channel in Graphics Mill for .NET

1. Check if alpha channel is supported. Not all the pixel formats support alpha channel. If you need to check whether current pixel format of the Bitmap class instance supports alpha channel, use HasAlpha property. If it returns true, alpha channel is available, otherwise - it is not.
2. Add alpha channel to the bitmap. To add alpha channel to the image, use AddAlpha(Single) method of the ChannelsProvider returned by the Bitmap. Alternatively you can convert bitmap to a necessary pixel format (e.g. Format32bppArgb) using Convert(PixelFormat) or ConvertToContinuous(ColorSpace, Boolean, Boolean) method of the ColorManagementProvider of the Bitmap.
3. Remove alpha channel from the bitmap. If you need to remove alpha channel, ChannelsProvider has a DiscardAlpha method. You can also convert bitmap to a certain pixel format (e.g. Format24bppRgb) using Convert(PixelFormat) or ConvertToContinuous(ColorSpace, Boolean, Boolean) method of the ColorManagementProvider returned through the Bitmap.ColorManagement property. Pay attention, when you discard alpha channel, you can either drop alpha without affecting color channels of the pixels, or flatten it by alpha blending each pixel with the specified color.
   
   For example, let's assume we have an image of ellipse with antialiased edges. Antialiasing is provided with variable alpha channel. If we just discard the alpha channel, we will lose the antialiasing:
   
   
   
   To get correct result, we should blend each pixel with some background color, e.g. white. For example, you can do it by passing color into DiscardAlpha method. You will get the following result:
   
   
4. Set some color transparent. You can emulate colorkeying by using Transparentize(Color, Single) method of the ChannelsProvider returned by the Bitmap. You can pass the color you want to make transparent into this method, and it recalculates alpha channel accordingly. Also, this method has the second parameter - tolerance. Tolerance is used to specify the "distance" between the colors to be interpreted as one. For example, RGB colors 0x000000 and 0x000001 look almost the same and you would expect the latter to transparentize too. But be careful with specifying high values of tolerance: in this case it may transparentize wrong colors.
5. Blend two bitmaps. To blend two bitmaps, you can use the Draw method of the Bitmap class with combine mode argument = Alpha. You can also specify overall opacity of the source bitmap.
   
   Alternatively you can use special transform class Combiner. It also has CombineMode and Opacity properties which work in the same way as the corresponding arguments of the Draw method. In fact, this class is used inside the Draw method.