BGRABitmap图像操作13：BGRABitmap的坐标系统

http://wiki.lazarus.freepascal.org/BGRABitmap_tutorial_13

    标准的画布采用整数坐标系统。CanvasBGRA 也是，不过反锯齿模式、alpha通道（透明）、gamma 纠正模式则不是。

  当使用纯整数坐标系统时，如果未启用反锯齿功能，坐标指定一个像素位置，即一个正方形。当画一条线，从(0,0) 到 (5,5)，左上方的像素是第一个画出的像素。标准画布中，最后一个像素没有画出，线的终点是(4,4)。

When drawing an ellipse, the bounding rectangle specifies the pixels that will be used to render the ellipse. Once again, in standard canvas, the lower-right coordinates are excluded from the drawing, so filling the rectangle (0,0)-(5,5) will in fact fill pixels having coordinates from 0 to 4.

    总之，不精确。

浮点坐标系统。

Now, when working with floating point values, the coordinates specify a position that can be anywhere on the pixels. The width value of the pen means really a distance, and not a number of pixels. So what does (0.0,0.0) mean ?

    浮点坐标则可以精确指定像素内的任何点。

Top-left distance floating point coordinates

It can be for example the distance from the top-left corner (this is not the case of BGRABitmap). In this case, this coordinate would be the upper-left corner of the first pixel. But if we do so, the behavior is slightely different between integer coordinates and floating point coordinates. Indeed, imagine we draw an horizontal segment on pixels (0,1) to (4,1). In pixel coordinates it would be the line (0,1)-(5,1) and the width would be 1 pixel. Now if we want to define this segment with floating point coordinates. The left side would be at 0.0, the right at 5.0, that's ok. The top side would be at 1.0 and the bottom at 2.0. Here is the problem. The center of the line is at a vertical coordinate of 1.5. So to draw this line, we should supply coordinates (0.0,1.5)-(5.0,1.5).

In fact every 1-pixel wide line with integer coordinates would be drawn between pixels, and with antialiasing, this leads to blurred lines. What appeared to be ok with horizontal coordinates is in fact an illusion, because if the line has caps, the correct coordinates are (0.5,1.5)-(4.5,1.5). If we ignore the last pixel issue, we see that these coordinates are simply 0.5 greater.

Pixel-center floating point coordinates

The coordinates can be the distance from the center of the top-left pixel. In other words, integer values are at the center of the pixels.This is the case of BGRABitmap functions. Using these coordinates, the line that fills pixels (0,1) to (4,1) is simply (0.0,1.0)-(4.0,1.0).

This is strange from a mathematical point of view, but very convenient, because you can use integer coordinates to draw normal 1-pixel wide lines. This way, there is little difference between calls to CanvasBGRA functions and BGRABitmap normal floating point functions.