光栅图像抗锯齿算法
我正在寻找一种有效的算法，该算法将查看四通道图像（RGBA），找到一个对象（一组 Alpha > 0 的像素，被 alpha = 0 的像素包围），然后尝试对对象的边缘进行抗锯齿。建议？
I am looking for an efficient algorithm that will look at a four channel image (RGBA), find an object (a group of pixels with Alpha >0 surrounded by pixels with alpha = 0) and then attempt to antialias the edges of the object. Suggestions?
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
我猜测您正在尝试描述一种情况，其中带有 alpha 的对象图像已粘贴到无 alpha 背景上，在 alpha > 的任何地方替换像素。 0，而不是正确混合。在生成的图像中，每个 alpha=0 的像素都是背景，每个 alpha > 的像素都是背景。 0 来自对象。没有混合，只是过度粘贴。
因此，一种猜测 alpha 现在非零的像素以前是什么颜色的算法可以解决这个问题。如果你有这个，你就可以在 0% < 的每个地方正确地进行混合。阿尔法< 100% 获得适当的抗锯齿边缘。
猜测缺失像素值的一种方法是使每个未知像素为其周围像素的平均值，并不断重复此操作，直到几乎没有变化为止。这几乎是使用附近像素修复旧照片图像中划痕的算法。您要求一种有效的方法。当存在较大的未知像素区域时，多重网格方法会让您更快地到达那里。
一旦你有了一个合理的未知背景，就用 alpha > 混合回像素。 0，这次正确考虑了它们的阿尔法。
I am guessing that you're trying to describe a situation where the object's image with alpha has been pasted onto an alphaless background, replacing pixels wherever it had alpha > 0, rather than blending properly. In the resulting image, every pixel with alpha=0 is background, every pixel with alpha > 0 is from the object. There's been no blending, just over pasting.
So - an algorithm for guessing what color the pixels where alpha is now non zero were before, would solve the problem. If you had that, you could do the blending properly, at every place where 0% < alpha < 100% to get a proper anti-aliased edge.
One way to guess the missing pixel values is to make each unknown pixel the average of the pixels around it, and keep repeating this until there is little change. This is pretty much the algorithm used for repairing scratches in old photographic images using pixels nearby. You asked for an efficient method. A multigrid method will get you there a little faster when there are larger unknown pixel areas.
Once you have a plausible unknown background, blend back in the pixels with alpha > 0, this time properly taking into account their alpha.