In order to solve the speckle noise residue and edge texture blur in phase image denoising, an adaptive total variation denoising method based on sine-cosine decomposition is proposed in this work. First, the original phase image is decomposed into two phase images by sine and cosine function. Then, the decomposed phase image is processed by adaptive total variation algorithm. Finally, the denoised phase image is synthesized by arctangent operation to quickly remove a large amount of speckle noise and retain the edge information of the image. The quantitative evaluation and analysis of the denoising results show that, compared with other denoising methods, the peak signal-to-noise ratio of the image obtained by this method is improved by 2.0 dB, and the structure similarity is higher. At the same time, the parameters can be adaptively selected to reduce the fluctuation of the phase image and improve the its quality.