分析图像的表示良好？
我可以使用一些算法来分析图像表示精度吗？诸如压缩算法设计人员之类的人是否具有比较两个图像表示形式的客观方法？
说我试图将圆形显示为栅格图像；分辨率越高，图像越接近一个完美的圆圈。随着您的前进，这些表示显然变得更加准确。
- ＆gt;
- ＆gt;
现在，我如何测量圆圈的特定表示与圆圈有多近？ 我想到的一种方法是测量不匹配的高res和低res图像（XOR）之间的位：
| 4.12％ | 1.15％ |
，但我将如何应用于非silhouette图像，例如照片或抗氧化图像？
Is there some algorithm that I can use to analyse image representation accuracies? Do people such as compression algorithm designers have some sort of objective way of comparing two image representations?
Say I'm trying to display a circle as a raster image; the higher the resolution, the closer the image comes to a perfect circle. The representations clearly become more accurate as you go along.
->
->
Now, how can I measure how close a particular representation of the circle is to the circle?
One method I came up with was to measure the area of the bits that didn't match between the high res and low res image (the XOR):
 |  |
|---|---|
| 4.12% | 1.15% |
But how would I apply this to a non-silhouette image such as a photo or an anti-aliased image?
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）
我认为您没有考虑镶嵌图像，这些图像易于从重复值的模式中检测到。
对于自然形象，问题没有意义。图像是尽可能准确的，可以执行区域采样（无论如何您都没有地面真相）。
这是您的抗血液图像：
I assume that you are not thinking of mosaic images, which are easy to detect from the pattern of repeated values.
For a natural image, the question does not make sense. The image is as accurate as it can, performing area sampling (and in any case you have no ground truth).
This is your antialiased image: