Is it necessary to do the needless work?

When we rotating, translating or scaling a scene, we must draw the scene again and again and it extremely slowing down the computer work, if the scene contains difficult images. Is there in OpenGL a way to avoid a reiteration?

This is how rendering applications work. You draw the whole scene every frame.

Best you can do is draw simplified geometry during the rotation/panning, and then the complete mesh again when the motion has finished.

Be sure to submit the geometry efficiently to the GPU. No need to transfer all this data each frame. Use at least old fashionned display lists, or even better : VBO (vertex buffer objects).

For zoom/pan, you may get away with capturing to texture the first render, then only manipulate a quad texture with it. Very fast.