General Idea:

For watermark like that can be removed by applying a mask over the image (most efficient way)

Recently I did a small project on topic like these, and you can find it here: https://github.com/ZingZing001/WaterMarkRemovalTool

The most important step in this process is that the program itself should be able to identify and remove the watermark correctly.

The logic and the pattern I used were:

Having two mask one through the RGB value and one through the HSV value

HSV Value:

If targeting a light-coloured watermark, the mask might have:

• Low Saturation (indicating less vibrant colours).
• High Value (indicating brightness).

And for RGB Value Mask: (To remove coloured watermarks)

• This mask identifies colors within the defined thresholds:
• R < 130, G < 200, B < 200: Likely targets darker colors with reddish tones.

A pixel is considered part of the watermark if it matches either the RGB or HSV masks.