Aiming at the problem of complex noise and fringe artifacts in current low-dose computed tomography (LDCT) reconstructed images, a LDCT denoising method based on dual attention mechanism and compound loss is proposed. This method obtains global feature information by introducing spatial attention mechanism and channel attention mechanism, and recalibrates the feature weights, so that important structural details can be retained, thereby improving the denoising performance of the network; at the same time, the perceptual loss measurement function is added to preserve the texture information sensitive to human eyes. Experimental results show that, in terms of visual effects, the proposed algorithm not only removes noise and artifacts in LDCT images, but also retains more texture features and structural details; objective indicators such as peak signal-to-noise ratio (PSNR) are are higher than that of other comparison methods.