Abstract: CMOS image sensor have the advantages of simple drive signals, single power supply voltage, high integration, low power consumption and strong radiation resistance. But in the field of space optical remote sensing, CMOS image sensor is not widely applied. The large-scale, high reading rate, high dynamic range CMOS image sensor is urgently in need. The 4 Megapixel CMOS image sensor of LUPA4000 is just such one. So LUPA4000 was chosen as the research target and its performance was tested. The image processing was conducted according to the results of performance test. In the research work, the parameters of defective pixels, response non uniformity(PRNU) and signal to noise ratio(SNR) were tested. The test results show that the CMOS image sensor of LUPA4000 has the following problems: the defective pixels are numerous, the PRNU is not ideal and the SNR is low. So the following image processing methods were adopted: dark background deduction to decrease the dark signal noise, defective pixels replacement to reduce the influence of defective pixels, and nonuniformity correction to reduce PRNU. The different processing combination among the three methods was applied .In order to evaluate the treatment effect of above processing combination, the contrasts of SNR and imaging quality between the original images and processed images were executed. The contrast results show that the method that combined defective pixels replacement with nonuniformity correction is the best combination.