c++ - 将胶片负 RGB 转换为正 RGB 的算法

Question

假设我有一张摄影底片扫描为 RGB 图像，我正试图找到一种算法将颜色值转换为 RGB 正片。

由于橙色偏差 (http://photo.net/learn/orange-negative-mask)，如果我简单地说 redPositive = 255 - redNegative，我得到的最终图像具有强烈的青色色调，并且非常褪色。这意味着这里给出的答案:Convert negative image to positive 不是正确的。

那么我将如何制作以下例程:

struct RGB
{
    unsigned byte red;
    unsigned byte green;
    unsigned byte blue;
};

void FilmNegativeToPositive(RGB const &negative, RGB &positive)
{
    // What goes here?
}

Answer 1

我没有数据可以测试，但根据你给的链接，底片是青色、品红色和黄色染料的混合物，不纯:

The yellow dye layer is the most pure. The magenta dye layer has a noticeable amount of yellow in it. The cyan dye layer has noticeable amounts of both yellow and magenta in it.

因此，您想做这样的事情(未经测试的伪代码):

Let I_MY be the ratio of yellow impurity to pure magenta dye
Let I_CY be the ratio of yellow impurity to pure cyan dye
Let I_CM be the ratio of magenta impurity to pure cyan dye

Given R, G, B in [0, 255]
Convert to CMY:
  C = 1.0 - R/255.0
  M1 = 1.0 - G/255.0
  Y1 = 1.0 - B/255.0

Calculate the impurities in the cyan dye and remove them, since we assume no other dye has cyan impurities:
  M = M1 - I_CM×C
  Y2 = Y1 - I_CY×C

Now the amount of magenta dye is correct, so subtract its yellow impurity:
  Y = Y2 - I_MY×M

Convert the corrected CMY values back to RGB:
  R' = 255×(1.0-C)
  G' = 255×(1.0-M)
  B' = 255×(1.0-Y)

如果事实证明存在比这更复杂的污染，你就会遇到线性代数问题:

[   1 I_MC I_YC]   [C']   [C]
[I_CM    1 I_YM] × [M'] = [M]
[I_CY I_MY    1]   [Y']   [Y]

在您要求解 C'、M' 和 Y' 的位置，然后转换回 RGB 颜色空间。