Aimed at the disadvantages of traditional threshold function in image denoising and the uncertainty of the threshold selection, a new improved threshold function algorithm in image denoising is proposed so as to further research on wavelet denoising algorithm. Matlab simulation of Circuit, Eight, and Road images are performed respectively by adopting the traditional hard threshold, the soft threshold, and the semisoft threshold method combined with this paper method. Through contrast and analysis, this method is good in visual effects and high in Peak Signal to Noise Ratio (PSNR). And at the same time, the signal distortion is solved in the soft threshold method, and the signal discontinuity and oscillation shortcomings in the hard threshold method are overcome as well. Moreover, the method can strike a good balance between noise elimination, and can also retain the weak features of the original signal and significantly address many problems in the traditional hard and soft threshold function denoising algorithms. This method is good in effectiveness in practical applications.