Abstract: Considering the human eyes- perception capability, the pseudo color coding technology is usually adopted to colorize the output gray image of infrared systems. Comparing to gray images, computers need more memory and calculation to process color images. Since some infrared equipments only provide pseudo color images,the grayscale conversion of pseudo color image was studied. Besides the 3 common algorithms, maximum decomposition, average and luminance, a decoding method was presented to decolorize infrared images. Based on the 2 kinds of methods above, decolorizing for both linear and sinusoidal nonlinear pseudo color images were studied in experiment; furthermore, image quality was also compared. The experiment result demonstrates that the decoding method is a kind of lossless algorithm for linear pseudo color image with 0 mean squared error. For sinusoidal nonlinear algorithm, loss due to quantization errors are inevitable, it still has a quite low mean square error (0.309 4) and high peak signal noise ratio (101.356 0).