In order to reduce the fixed pattern nise (FPN) induced by dark current in CMOS image sensor,this paper presents a real-time dark current calibration method ba sed on image processing technology.This method is implemented and integrated int o a 4T CMOS image sensor which is serving as a 50f/s video camera.In this pape r,the composition of the 4T CMOS image sensor′s ADC output is analyzed.Accordin g to the analysis,the dependences of the dark current induced FPN on exposure ti me and temperature are modeled.After the black level is obtained and subtracted from the sensor′s ADC output,the dark current factor matrix is calculated by su btracting reference pixels′ (exposure time of 0) outputs from active pixels out puts.The dark current factor matrix is calculated every 10° centigrade and save d in a flash memory.By interpolation,the dark current factor matirx can cover th e 4T COMS image sensor′s working temperature range.Finally,based on the 4T CMOS image sensor′s exposure time and gain settings,the proper dark current matrix is selected and read in via SPI.Then the dark current induced FPN is evaluated a nd reduced proportionally.The experimental results reveal that by using this met hod the dark current FPN is reduced by more than 75% with the CMOS image sensor ′s working temperature range,and the accurate black level is also obtained.