In view of the data frame obtained by the mobile phone CMOS camera which is regard as the receiver in VLC link

the data frame has a low extinction ration due to the blooming effect of the CMOS.In addition to the background light

the thermal noise of CMOS will reduce the data frame recognition rate.Aiming at the above problems

an optimization method using high-low-hat transformation and maximum value suppression is proposed.This method improves the extinction ratio of the data frame and suppresses the local noise.Therefore

the data frame can be decoded easily.However

the optimization algorithm is time consuming.In order to transplant the algorithm into the mobile phone

the data frame is processed by a short time-Local Weighted Scatterplot Smoothing（lowess）signal recognition method.The results show that the lowess algorithm is less affected by local noise

and the optimal signal to noise ratio can be obtained when compared with previous algorithm.