图是最好的解释，你不明白么？
镜头的任何技术、特性，最终都必须反映在照片上（即“图”），照片是最终的结果，眼睛是最后的裁判，这个道理都明白吧？
当眼睛无法区分的时候才需要借助于仪器和软件，如果和视觉方向相悖，则应根据人眼视觉予以修正。
分辨率本身就是反差的度量，反差就是对比，改善了对比就是提高了分辨率。 本帖最后由 一碗水不平6 于 2014-4-18 22:35 编辑

你在引用Slrgear边角对比照片的时候，没看关于这张照片的文字论述吗？我把这段文字粘过来了，如果你确实没看的话，请认真看一遍再发言吧。

Micro-contrast
In case you were thinking we wouldn't get around to discussing the micro contrast beyond the above, fear not!

Given Kazuto Yamaki's decision to prioritize micro contrast at the expense of sharpness, we had to take a look at this aspect of performance. The results were amazing.

You see, while the resolving power of the Otus technically barely nosed out the Sigma in our blur testing, the perceived sharpness of the Sigma is actually greater than the Zeiss lens. For the most dramatic example of this, take a look at this crop from the upper left corner (mouse over the image to see the crop from the Zeiss 55mm f/1.4 Otus lens):

How can this be? Well, resolution can be defined as the ability to differentiate between closely-spaced transitions from light to dark. The Otus clearly has this in spades.

But the quality of these transitions is important for our perception of sharpness; sharpness and resolution are related to each other, but it's possible for a lower resolution image to actually look sharper to our eyes. Looking at the alternating light/dark lines in the Sigma shot, you can see that it clearly renders them with crisper edges and higher contrast between the light and dark areas.

The other factor at play here is the CA (and perhaps a bit of coma distortion and general veiling flare in the corners), which are poorer on the Sigma than the Otus. This reduced contrast and lightening of the dark target elements where the "purple haze" intrudes on them is interpreted by DxO Analyzer as blur. (Which it is really, just with a different characteristic than the diffuse blur we commonly refer to with that term.) So, while it is technically blur, the image still looks sharper to our eyes than that of the Zeiss, as you can see from the Sigma crop above.

When comparing the blur graphs for the two lenses, another point comes into play, namely that the DxO Analyzer software we're using here measures blur at 17 points across the image, which our graphing software then draws smooth contours to fit. This gives a pretty complete and easily digestible view of a lens's blur characteristics, but we're not seeing every single point in the frame. Within the bulk of the frame, it's unlikely that a lens will do something crazy entirely between measurement points, such that it won't show up in the plots. However, it is possible (as is the case with Zeiss 55/1.4 Otus), that a lens might turn bad at the very edge of its image circle, and not be picked up, because the behavior is outside the measurement area. At the corner-most locations that DxO Measured, the Zeiss settled down pretty well, and is actually a bit better looking than the Sigma, but as you can see from the crops above, the Sigma looks noticeably better in the extreme corners. (And its geometric distortion there is significantly better as well.) 本帖最后由 longyu1129 于 2014-4-18 22:10 编辑

那只能说明软件有问题吧？软件只应该提高精度，不应方向和人眼相反。
难道拍照觉得模糊但心里告诉自己：我这个看着模糊其实分辨率更高，只是肉眼看起来模糊，实际上更清晰么？

我还真认真看了一下，大概意思是适马使用了一种新技术，在传统认为“分辨率较低”的情况下可以获得更高的清晰度，所谓微对比度。根据我的认识，如果属实，只能说明用传统“分辨率”来作为衡量镜头的分辨能力是不足的，不全面的，不足以表示镜头真正的光学素质，和人眼认识不一致，应予以修正。 本帖最后由 一碗水不平6 于 2014-4-18 21:24 编辑

你是肉眼观察得出的结论，DXO依靠机器和软件判读，和人肉眼感受不同。你引用的Slrgear也在review里明确指出Sigma大光圈下像场平整度和Otus差很多；就在你引用的那个边角对比图里，他们也就反差和肉眼感受的问题做了分析和解释，这个结论和DXO是一致的。你认真看一下吧。

中心S50A稍锐，这和DXO的一致，但但我明显看到边缘也是S50A锐：
边缘图中那个竖向最小的三个小黑条S50A能分清，Otus已经糊在一起了，这还要说Otus边缘比S50A好得多，这不瞎掰嘛！
难道是DXO的算法太奇葩的缘故？

DXO更多依靠机器判读，有其一套固定算法，是很严谨地。Sigma的反差可能高一点，因此细节更容易被人肉眼识别，这和机器判读是有区别，Slrgear的评测里对这一点有明确论述；Slrgear自家的结论也是大光圈下，Otus的像场平整度要显著优于Sigma。当然，在实际拍摄中，那个更好，就很难说了，估计分不出高下。

一块比，DXO牛大了

DXO的数据难道是像马航客机坠毁那样通过“计算推理”得到的？

再看看DXO的数据，笑死人……

实拍数据来源：slrgear.com
http://slrgear.com/reviews/showproduct.php/product/1677/cat/30
以下内容由 一碗水不平OO 于 2014-4-18 17:03 补充
DXO的数据难道是像马航客机坠毁那样通过“计算推理”得到的？ 以下内容由 一碗水不平OO 于 2014-4-19 00:17 补充
找到为什么DXO将更清晰的S50A边角解释为边角分辨率差的原因了，看slrgear.com的原文说明：
“The other factor at play here is the CA (and perhaps a bit of coma distortion and general veiling flare in the corners), which are poorer on the Sigma than the Otus. This reduced contrast and lightening of the dark target elements where the "purple haze" intrudes on them is interpreted by DxO Analyzer as blur. (Which it is really, just with a different characteristic than the diffuse blur we commonly refer to with that term.) So, while it is technically blur, the image still looks sharper to our eyes than that of the Zeiss, as you can see from the Sigma crop above.”
——果然，是DXO那个垃圾分分析仪的问题，S50A全开光圈边角轻微的色散被解释为模糊，使得S50A得分边角大大低于Otus，实际上，眼睛看S50A明显清晰很多，仪器弱智，用它来做标准？
就好比，用仪器测两个人（A重80kg、B重85kg）比谁更重，仪器发现B身高更矮，于是将其解释为体重轻——真是搞笑啊！
以下内容由 一碗水不平OO 于 2014-4-19 00:18 补充 以下内容由 一碗水不平OO 于 2014-4-19 11:26 补充
有色友说去色后会有不同，我试了，去色后无论中心还是边角，仍然是S50A好，看图：