Traditional progressive strategy for denoising cascades a series of backbone denoisers to enhance the performance. However for denoising task with low peak signal-to-noise ratio(PSNR), we find this strategy is ineffective. Thus we extend the traditional progressive strategy to nonlinear progressive strategy learning to dig non-noise component from discarded noise of backbone denoisers. Inspired by the workflow of archaeology, the proposed strategy alternatively and repeatedly implement backbone denoiser and non-noise component digging module in a progressive manner. For traditional and deep denoisers, experiments show that for low PSNR images with regular shapes, the proposed strategy is able to help backbone denoisers recover these shapes with better discriminability than traditional progressive strategy. Although experiments find the proposed strategy help them achieve better performance on several public datasets with clear-cut rules, we make no claim that these published methods accompanied by our strategy will beat the state-of-the-art current algorithms on these and other natural image datasets. The novelty is that the proposed strategy is general and interpretable which can be applied to various deep or traditional denoisers for stronger nonlinear fitting capability and reliable performance improvement on severely ill-posed low PSNR noise removal problem.