A scene-by-scene (S×S) compression process enables improved visual quality within applicable compression constraints to be obtained. A compressed video signal is analyzed, generating compression statistics. A scene of the video signal is identified for re-compression and a maximum target bit rate for re-compressing the identified scene is determined based on the compression statistics and applicable compression constraints. A target bit rate for re-compressing the scene is selected based on the maximum target bit rate, and compression parameters for the scene are determined based on the scene's compression statistics and the selected target bit rate.