Not all colors are displayed correctly in DICOM PR graphics

[milkyway007]

Describe the bug
I am working on a piece of software that generates DICOM Color Softcopy Presentation State files. I am testing produced CSPS files using the Weasis viewer. Everything works fine. However, some colors are distorted. Green (0,128,0), blue (0,0,255), black (0,0,0) color is displayed correctly. Yellow (255,255,0) is changed to light green (~99,255,0), white (255,255,255) - to light blue (~99,255,255), red (255,0,0) - to dark red (~99,0,0), orange (255,165,0) - to light green (~99,165,0). As you can see, the problem is with the red channel. Its value is multiplied by approximately 0,38, regardless of concrete value (I have tested different values). I am not able to figure out problems with our software.

Desktop (please complete the following information):

• OS: Windows 10
• Java Rutime: OpenJDK Runtime Environment AdoptOpenJDK (build 14+36)
• Weasis Version: 3.6.1

Screenshots
Original image

The result displayed with Weasis

[nroduit]

This explanation does not allow me to draw any conclusions.

If you are waiting for an answer, you should provide a more precise explanation about the DICOM attributes or an example file.

[milkyway007]

I will attach a zip with
secondary capture image and color softcopy presentation state files.
DICOM_files.zip
Inform if you need more info.

[nroduit]

You store RGB value into an attribute requiring a CIELab Value

https://dicom.innolitics.com/ciods/pseudo-color-softcopy-presentation-state/graphic-annotation/00700001/00700209/00700233/00700251

[milkyway007]

Thank you very much for your answer.

Based on it I wrote an RGB to CIELab converter. The source code is attached to my answer. Firstly, this class converts an RGB value to a CIELab value. Then L, a, and b values are linearly scaled to 16 bits as the standard requires.

The results are as follows:

Color
Green
RGB: 0,128,0
CIELab: L: 46,228760110623; a: -51,6991318531787;b: 49,8975342624086
Scaled CIELab: 30296\19609\45720

Blue
RGB: 0,0,255
CIELab: L: 32,3025866672495; a: 79,1966617893093;b: -107,856977530782
Scaled CIELab: 21170\53250\5177

Black
RGB: 0,0,0
CIELab: L: 0; a: 0;b: 0
Scaled CIELab: 0\32896\32896

Yellow
RGB: 255,255,0
CIELab: L: 97,1382469812973; a: -21,5559083348323;b: 94,4824854464446
Scaled CIELab: 63660\27356\57178

White
RGB: 255,255,255
CIELab: L: 100; a: 0,00526049995830391;b: -0,00428585423328798
Scaled CIELab: 65535\32897\32895

Red
RGB: 255,0,0
CIELab: L: 53,2328817858425; a: 80,109309529822;b: 67,2200683102643
Scaled CIELab: 34886\53484\50172

Orange
RGB: 255,165,0
CIELab: L: 74,9322155874529; a: 23,9358747104639;b: 78,9565229112003
Scaled CIELab: 49107\39048\53188

Using this thing, for example, http://colorizer.org/ you can control that produced CIELab values are correct.

Produced DICOM images are located in the attached ZIP.

Weasis screenshot

So, Weasis shows something very different from the original image.

ToCieLabConverter.txt
DICOM_files.zip

It seems that Weasis expects more likely RGB values but not always reads them correctly.

[nroduit]

As this is the first time I see a DICOM PR using the PatternOnColorCielabValue field, I realize that the implementation was messy. The examples I had were using GraphicLayerRecommendedDisplayRGBValue.

I remade the implementation (in the last 2 commits) and now it displays properly.

[milkyway007]

Thank you for your cooperation!