In order to get more accurate placement information of radiation thera py and reduce the radiation dose received by clinical patients in the hospital,a new denoising algorithm based on the noi ses in the cone-beam computed tomography (CBCT) projection data is propo sed in this paper.Firstly,the appropriate value of the filter strength is determin ed by the noise standard deviation,the edge information and the average of the average gradient values of texture sub-block s which are calculated in different angles of the projection data received before the reconstruction of the 3D CBCT images,th en the projection data are denoised by the adaptive nonlocal means (NLM) algorithm and lastly,the denoised projection data are reconstructed to obtain high quality CBCT images.In addition,a noise estimation algorithm is improved in order to make i t more suitable for the CBCT projection data.The experimental results show that the proposed algorithm can estimate the noise level of the projection data,the new algorithm is also superior to other methods,it can reduce the noises,keep the details and enhance the contrast of the images effectively,so the proposed algorithm is beneficial to acquiring the accurate placement information and diagnosing in the clinical practicefor the do ctor.