A method is proposed of selection, for a current image portion to be encoded and for at least one encoding module included in a video encoder, of at least one encoding parameter from amongst a set of encoding parameters available for the at least one encoding module. The method is based on a decision model defining a minimization, under a rate constraint, of a first measurement of distortion between source samples, included in the current image portion to be encoded, and current reconstructed samples, included in a current reconstructed image portion, obtained from the current image portion to be encoded. The decision model defines the minimization under an additional smoothness constraint, pertaining to a second measurement of distortion between the current reconstructed samples and preceding reconstructed samples, belonging to a temporal reference neighborhood comprising at least one preceding reconstructed image portion obtained from at least one preceding encoded image portion.