Gait is an important physiological biometric in the area of computer vision for human authentication at a distance. In appearance-based gait recognition system, significant gait features could be affected by various cofactors such as cloths or carrying objects. Therefore, detecting co-factored segments and eliminating co-factored information without losing the features of Gait Energy Image (GEI) is one of the major concerns for appropriate gait recognition. In this paper, we proposed a method for detecting cofactor affected segments of GEI and an approach for dynamic reconstruction of co-factored GEI for more accurate gait recognition. The whole GEI is first segmented into three parts considering the area of cofactor appearance in it. Moreover, co-factored information are detected and eliminated depending on some predefined threshold values. Finally, the three segments are recombined for final classification. The CASIA gait database is used here as a training and a test data. The result shows better performance with 85.04% accuracy which is more convenient than other conventional gait recognition methods.