False color suppression problem


Hello, rawtherapee team!
Recently, I am looking for a raw image processing program that support linear base curve. I found both rawtherapee and darktable are good candidates for me. Through comparison, I think rawtherapee has better compatibility, because it supports dcp and lcp profiles, while darktable doesn’t. So I want to stick with rawtherapee.
However, I found some problem in False color suppression (FCS) option. In short, it grays the color too much compared to the color smoothing (CS) option in darktable.

Here are some comparisons, both using AMaZE method, and have disabled chromatic aberration, defringe, and lens correction:

  1. rawtherapee (FCS = 0) vs. darktable (CS = 0)
    (Because only 4 images can be uploaded in one topic, I will try to upload them in the reply)

  2. rawtherapee (FCS = 5) vs. darktable (color smoothing (CS) = 5)

Without FCS and CS, rawtherapee and darktable nearly show the same results, with some false colors near the edges of the text. The false colors are well suppressed with FCS = CS = 5.
However, in the sense of pokers, FCS = 5 make rawtherapee grays more, especially in some narrow or small color region, while CS = 5 in darktable still keep the saturation well.

I am not familiar with demosaicing and FCS algorithm. Are there some parameters like suppression radius, threshold, or strength which are not shown in the user interface? I am looking forward to some improvement in FCS.

Here is the RAW file:

Best wishes!

  1. rawtherapee (FCS = 0) vs. darktable (CS = 0)

(Morgan Hardwood) #3

Hi @Chlheng

  1. Have you only tested FCS=5? If it’s too strong, have you tested lower values?
  2. Have you tested other demosaicing algorithms, raw CA correction, and defringe?

  1. I have tested FCS = 1, 2, 3, 4, 5. If FCS is low, the false color suppression is too weak for me. Beside, even FCS = 1, the color saturation in some narrow and small color region still lower than that in darktable with CS = 5.
  2. I have tested all demosaicing method, the IGV and IMMSE method produces few false color, but show dot like pattern (like bayer pattern) in some region in the poker sense. The AMaZE method gives the best overall quality for me. Otherwise, defringe has no effect on false color in the text sence, and CA will loss some detail.

(Morgan Hardwood) #5

Confirmed, I tested using that raw file and found no perfect all-round solution.


Thank you for your confirmation. Good luck!