Abstract: The selection of threshold plays a crucial role in denoising process for an image using wavelet transform.To remedy drawbacks identified in the traditional thresholding techniques，this paper presents an improved method of threshold quantization with continuity and differentiability，as well as a threshold calculation method with self-adaptability.The continuity and differentiability are achieved by using cubic polynomial interpolation based on hard thresholding.According to the Birge-Massart standard，the threshold is calculated at each decomposed layer.As a result，the threshold can achieve the local self-adaptation.The experimental results show that the proposed method has both better subjective visual effect and the PSNR than the traditional approaches.

Key words: wavelet denoising, threshold, cubic polynomial, continuity, self-adaptability

摘要： 阈值的选取是小波域图像去噪的关键技术之一，但传统的阈值各有其缺陷。提出了一种连续可导的阈值量化方法和具有自适应性的阈值计算方法。用三次多项式在硬阈值的基础上插值，使新的阈值函数保持了连续性和可导性。按Birge-Massart规则分层计算阈值，使阈值具有良好的局部自适应性。通过实验表明，该方法去噪后的图像主观视觉效果和峰值信噪比均比传统算法优越。

关键词: 小波去噪, 阈值, 三次多项式, 连续性, 自适应