Facial deception detection has become a popular and challenging problem. In this study, an effective system is proposed to address this issue based on visual clues. The Parametric-Oriented Histogram Equalization (POHE) is presented to enhance image contrast and reduce the noise effect. A random forest classifier is applied to track the facial landmark points, and they are subsequently utilized to analyze the facial action unit based on the movement of the facial feature points. In addition, the geometrical features are also considered, and then the Sequential Forward Floating Selection (SFFS) is integrated to select the best feature combinations. To verify the extracted features for deception and truth identification, the Support Vector Machine (SVM) is applied. Experimental results demonstrate that even under uncontrolled factors, e.g., illumination, head pose, and facial sheltering, the proposed method is consistent in achieving an effective recognition results and provides superior performance than that of the state-of-the-art methods.