帮分析下图像远处天空偏红的原因

sea_83

白平衡调好后,去测试发现如果光圈加大/曝光时间增加,使光通量增加,远处的天空还有一些较明亮的地方会偏红,而其它部分颜色看着很好.
白平衡已经测试了很长时间,虽然红色分量的增益设的比较大点,但对着色板还不错的,
另外滤色片用的是400~700nm的不知道会不会因为红色过高引起这个现象
有朋友说可能是色温引起的,不知道会不会是这个原因
另外我做yuv转换后对yuv分量都乘了1.2左右的系数,这个会影响么?
[upload=jpg]UploadFile/2008-2/08226@52RD_未标题-1.jpg[/upload]

coma

车里的人好酷阿[em02]

vasdf

滤色片搞到650看看

edmt1030

1. lens is no problem?
2. IR-Cut filter is used?
    Is it really get for 400~700nm by using this filter you used?
3. Make sure the AWB operation range which the device can support you used now[br]<p align=right><font color=red>+1 RD币</font></p>

sea_83

回楼上，镜头在其它几个产品上用过应该没问题,(如果有问题,会是什么原因?镜头也有滤色的作用?)

滤色片是定制的400~700nm,都已经加上了

从图像数据看,亮的区域都没有饱和,只是红色分量偏大

edmt1030

can you take a picture when shooting object indoors?
compare the image indoors and outdoors is one tool for debug..

if this lens can work well on the other board, lens problem possibility becomes low..

if spectrum curve is fixed within 400-700nm, please try this filter on the other different design board, if the same phenomenon occurs, this should not be filter issue...

because this red color is not only appearing in the sky area but also the road area, so I am doubt that AWB is really no problem?[br]<p align=right><font color=red>+3 RD币</font></p>

chengsean

You don't do minimum gain calibration.
The R, and B channel is saturation when G is not saturation.

This problem is not related to spectrum problem just sensor response problem.

jimi

看图片的直方图, G B channel 在右端出现山峰形状，出现了饱和现象，问题出在G B channel 输出范围（正常为 0-255)不足导致亮部偏色，可能是自行修改了一些图像转换系数导致。
[upload=jpg]UploadFile/2008-2/08228@52RD_123.jpg[/upload]

sea_83

<P>谢谢楼上的热心帮助和分析
补一张夜间的图片,这是辆白车,有打闪光灯
从图片看在暗的区域R&lt;B,但渐变亮的过程中R增加的幅度明显高于B,直到特别亮的地方R&gt;B  但G分量一直偏小
AWB可能跟edmt1030说的一样有问题.测试了很久想找到个随光强变化影响不的平衡点真的很难</P>
<P>估计在做AWB算法调节的时候R B G之间的增益差别不能太大,有经验的能否提供点建议,谢谢
<B>[upload=jpg]UploadFile/2008-2/08228@52RD_2.jpg[/upload]</B></P>

hejun0323

是不是使用的红外镜头。可以对RGB整体调整，将低级灰阶局部调高，是整体显得比较苍白，就会好点了啊

jimi

不是镜头和IR的问题,也不是AWB的问题,而是你自己改了一些图像转换系数如RGB>YUV等造成的.

chengsean

[upload=jpg]UploadFile/2008-2/08228@52RD_minimum Gain.JPG[/upload][br]<p align=right><font color=red>+3 RD币</font></p>

hejun0323

版主能不能具体解释一下为什么图片输出的转换会造成偏色啊？那需要对那个分量进行调整啊？单独调整S分量会不会有改善啊（把S分量调高）？

jimi

hejun0323:先把所有改过的图像处理参数重置为default, 一步步分析哪项设置是错误的原因。

jimi

<DIV class=quote><B>以下是引用<I>chengsean</I>在2008-2-28 23:09:46的发言：</B>
[upload=jpg]UploadFile/2008-2/08228@52RD_minimum Gain.JPG[/upload]</DIV>


这样的示意图是否更合适？
[upload=jpg]UploadFile/2008-2/08229@52RD_cccc.jpg[/upload]

ly8211

颜色格式 转换的问题。就如版主说的那样。 查查正确的转换算法。

chengsean

<DIV class=quote><B>以下是引用<I>jimi</I>在2008-2-29 9:38:07的发言：</B>



这样的示意图是否更合适？
[upload=jpg]UploadFile/2008-2/08229@52RD_cccc.jpg[/upload]</DIV>


What I mean is the linearity of the R, G, B channel.
The AWB function is used to adjust the slope of R, G, B channel.
So after you do AWB the R, G, B channel slope of the neutral color will be the same.
If your linearity of the sensor response is good, this scene will not apperar after AWB.
Because we always adjust the slope to the Max channel value (such as G in most sensor).
The graph I show is the bad linearity of G channel . If your linearity of the R, G, B
channel is as what I show. When the pixel value is close to saturation. You will
multiply the gain on the R , G, B channel. But G is not linear close to saturation, you
can't multiply the gain on it.
For example when the pixel value of R, G, B channel in linear is
R = 100, G = 130, B = 100; And the R, G, B channel in non-linear is
R = 200, G = 240, B = 200; the R Gain = 1.3,  B Gain = 1,3.
After AWB, The linear value will be R = G = B = 130., but non-linear part will be
R= 255, G=  240,  B= 255; But In theoretically, the G value should be 255 (If the channel is linear)  <FONT color=#ff0000>+3RD币</FONT>

jimi

你的意见是对的，白平衡是有作用的，是通道的最大值不足导致的偏色

jimi

这个图会比较近似
[upload=jpg]UploadFile/2008-3/0833@52RD_cccc.jpg[/upload]

mylovefor

谁有MT9D111的开发资料或数据手册,可不可以给一下......
right_guy@163.com
[em03][em03][em03][em01][em01]