Computed tomography reconstruction methods consider a normalized input signal. Images must be corrected using both a flat field record and a dark field record. Whereas there is no physical reason why the dark field would evolve during the experiment, the flat field may vary with time. One possible cause for flat field variation is irradiation damage which indeed occurs on the camera considered in this study. This paper considers different methods for taking into account the evolution of flat field due to irradiation damage. The aim is to keep the error resulting from bias on flat field at the same level than signal to noise ratio on fully illuminated areas of images, that is to say an error of 1%. Though this is still a work in progress, the methods proposed will be discussed and compared.