Terahertz wave imaging has many advantages such as small ionizing radiation and high discrimination of organic matter, so it is very suitable for non-contact, non-destructive imaging detection. However, this imaging method is susceptible to unavoidableproblems such as disturbance of electromagnetic environment and fluctuation ofequipment power, which makes it difficult to obtain high resolution terahertzimages. In this paper, terahertz imaging is realized by using the terahertz continuous-wave source imaging system. The terahertz images of four samples of cards, leaves, coins and keys are denoised by ourimproved threshold gray-scale transformation algorithm. Then, the images are sharpened and enhanced by the Laplacian operator. The results are evaluated bymean square error and peak signal to noise ratio estimation. After the denoising, the estimated peak signal to noise ratio of some samples increased by 4～5 dB, so the image quality was improved obviously. This research proves that the built-up terahertz imaging system has a pomosing application prospect.