研究X3元件 每层的细节损失
1061
18
|
[1 楼] 418588326
[泡菜]
14-2-11 20:17
新的DPQ元件公布。很多人讨论说 老的X3元件只有第一层蓝色是高分辨率的 后面每一层都会有损失
为此特地开贴用眼睛看结果吧。 |
|
[19 楼] 可爱的卷卷
[泡菜]
14-2-12 11:46
亮度+色彩1+色彩2。是个不错的想法,PS中LAB之强大可见一斑。度度层可以理解成一张黑白照片,锐度主要和这个层有关,后两层负责上色,所以2000W相素的黑白片比1500W相素的彩色片锐度更差,这点不会成立。当然也就是说在不插值的的情况下,量后输出相素也就只有2000W。也比DP-M的1500W左右好。而色彩上LAB相对RGB更有调色的优势,所以理想的情况下,适马把可能是想通过把N年没有调好白平衡的X3,通过LAB方式解决问题。当然不理想的情况下是会不会越调越偏谁也不知道,只有一点是肯定的,适马想通过换个角度看问题来解决问题。希望成功,自从用过DP1S,对适马机身感觉不错,虽然后来出掉了。正准备14年搞个DPM结果这DPQ出来了,还是观望一下再说。说不定真解决了N年来的老问题。
|
|
[18 楼] qingjun
[资深泡菜]
14-2-12 11:19
|
|
[17 楼] kusoking
[泡菜]
14-2-12 10:17
样张说明一切
|
|
[16 楼] 城管又滚了
[注销用户]
14-2-12 08:49
用户已注销,历史内容不予显示
|
|
[15 楼] siliconworm
[资深泡菜]
14-2-12 01:05
xieqiao 发表于 2014-2-11 20:39 这个模拟是靠谱的,理论上讲2000W+490W+490W三块感应器要达到和原来真正的X3相同的锐度,实际分辨率就只有490万像素,适马这不是在进步,是在削减成本坑爹啊! |
|
[14 楼] 江春入旧年
[泡菜]
14-2-11 22:09
|
|
[13 楼] 该人员
[资深泡菜]
14-2-11 22:03
|
|
[12 楼] 418588326
[泡菜]
14-2-11 21:10
|
|
[11 楼] 418588326
[泡菜]
14-2-11 20:58
|
|
[10 楼] xieqiao
[泡菜]
14-2-11 20:39
国外有网友用DP3M图像模拟演示,左边是原来图像,右面是模拟dp Q的。
http://www.dpreview.com/forums/thread/3621631#forum-post-53085525 Quattro simulation To be honest, I am a bit skeptical about the new Sensor design in terms of resolution, though I´m very curious to see actual images. I just tried to simulate the resolution, using a crop from an DP3M image. Here´s what I did: I copied the green and red channel in PS separately, reduced it to 50% size, enlarged it back to original resolution and replaced the respective channels in the original image. I know, this isn´t quite what Sigma will probably be able to do with the RAW data, but maybe it illustrates what happens to the overall image resolution. I´m not quite sure if I like the result. But maybe, I am completely wrong about this ... left is original, right is simulated In reply to Guenter Borgemeister, The simple B / G / R explanation is purely for marketing reasons. This means that your simulation does not come in any way close to what the quattro sensor does. In order to do that you would need to do the following four steps: 1. B / G / R to R + G + B / R + G / R 2. summing of 4 neighbouring pixels' R + G and R channel values 3. upsamling of R + G and R channels 4. backtransform R + G + B / R + G / R to B / G / R. Greetings, Robert F. Tobler ~~~~~~~~~~~~~~~~~~~~~~ In reply to Robert F. Tobler, Robert F. Tobler wrote: '1. There are no "blue green red" in Foveon. The layers are Ψ1 = f1(Luminosity) + φ1(Blue) Ψ2 = f2(Luminosity) + φ2(Yellow) Ψ3 = f3(Luminosity) + φ3(Magenta) 2. Two bottom layers in the X3 arrangement are not sharp, especially the bottom one.' 本帖最后由 xieqiao 于 2014-2-11 20:44 编辑 |
|
[9 楼] 该人员
[资深泡菜]
14-2-11 20:38
记得佳能论坛有个spot论证过梅林X3红蓝色分辨率大大高于1600万的msk,但是在有足够绿色信息情况下梅林X3的分辨率大致相当于无低通的1600msk。那个贴子里有老x3和msk在各种颜色条件下的分辨率测试数据。可以发现老x3黑白分辨率和各种彩色分辨率测试数据全都一致,msk的分辨率则随绿色信息的多少而变化。这是不是可以证明x3二三层的分辨率衰减不严重呢?
|
|
[8 楼] lyticast
[泡菜]
14-2-11 20:30
|
|
[7 楼] moreyoung
[老坛泡菜]
14-2-11 20:29
这颜色是怎么分离出来的
|
|
[6 楼] 418588326
[泡菜]
14-2-11 20:23
恕我狗眼 没看出三个通道解析度上有很大区别
|
|
[5 楼] 418588326
[泡菜]
14-2-11 20:22
红色
|
|
[4 楼] 418588326
[泡菜]
14-2-11 20:21
绿色 300%放大
|
|
[3 楼] 418588326
[泡菜]
14-2-11 20:20
用的是SD1 镜头是50 F8
蓝色通道 单独输出 然后300%放大 比较便于观察 |
|
[2 楼] 418588326
[泡菜]
14-2-11 20:18
100%放大
|
