Yes—pixel count alone does not tell you measurement precision. What matters is how large the subject projects onto the sensor, which depends on subject size, distance, focal length, and sensor size.

Your approach is basically correct for a simplified estimate:

1. Compute the subject’s image size on the sensor from geometry.
2. Convert that sensor size to pixels using sensor width/height and image resolution.
3. Divide real subject size by pixel count to get mm per pixel.

In your example, if 100 mm of subject width spans about 1901 pixels, then the sampling is about 100 / 1901 ≈ 0.053 mm per pixel, which suggests 0.1 mm resolution is plausible in theory.

But practical accuracy is worse than this ideal estimate unless:

• the object lies in a single flat plane,
• that plane is perpendicular to the camera,
• focus is sharp,
• lens distortion is calibrated,
• the true focal length / geometry is known.

This is essentially a photogrammetry problem. For reliable measurements, camera calibration is important. Also remember that real resolving power is lower than raw pixel count suggests because of optics, blur, and sensor limitations. So use your pixel calculation as a first-pass screening tool, not a guarantee of 0.1 mm measurement accuracy.