Face expression recognition has made considerable growth in last decade due to its versatile applications in various fields such as emotion analysis, biometrics, surveillance and robotics. This paper proposes a quality face descriptor in which the local information of the face surface is concealed by means of Local Directional Number (LDN) pattern and it generates a more discriminative code compared to current methods, with the aid of compass masking it computes the arrangement of each micro pattern by which the local information is take out from the identical facial patterns that have contrasting intensity alteration. Then the face is divided into many sections, from which LDN components are extracted, followed by integration of these components and it is used as a face descriptor. Considerable tests were conducted on it under illumination and different facial expressions to produce truthful face expression recognition with better recognition rate.