It’s onerous to know what is going on on with out photographs or code so unsure if this addresses your specific concern or not, however ‘Linear’ on this context implies that OpenGL applies a bilinear filter to the feel when it’s sampled.

A bilinear filter is a median of the pixels from the unique picture across the pattern place so you could get barely blurry outcome relying on the picture’s unique measurement and the way it’s scaled.

If you do not need your textures to be filtered in that approach, it is best to apply the Nearest texture filter mode which simply picks whichever pixel from the unique picture is closest to the requested pattern place and attracts that as an alternative of making an attempt to be intelligent about it. (This is what provides video games like Minecraft their distinctive ‘chunky pixel’ look in its textures)

Here’s the distinction on a small picture that has been scaled up:

(nearest neighbor, bilinear, bicubic [not directly supported in OpenGL])

Hope that helps!