Dear OpenCV Community,
I am encountering an issue with cv2.matchTemplate when performing template matching. Specifically, colored templates are being matched on black-and-white images, and vice versa, with a similarity percentage of 90%. This behavior is undesirable, and I want to ensure that such matches do not occur. Could you please guide me on how to avoid this in my module?
Example
I used an image containing three rectangles of different colors as the target. The template was a cropped region containing one rectangle. Despite the rectangles having different colors, template matching still reported a 90% similarity when comparing the colored template with differently colored regions in the target image. Both images had the same size and shape.
Experiments
-
Color-Based Comparison: I attempted to compare the template and the target image in their color form (e.g., RGB).
-
Channel Splitting: I split the color channels of both the template and the target image and performed template matching on individual channels.
Even after conducting these experiments, I faced the same issue: templates of different colors still match with high similarity scores in the target image.
Could you suggest an effective approach to ensure the module recognizes and respects the color difference during template matching?