Under the Internet of things, the intelligent visual image is a high noise image. Because the fixed threshold (or block to get the threshold) used in the general fixed threshold segmentation and adaptive threshold in combination with the wavelet denoising algorithm can not achieve the target location when the transition “interference” between the targets to be segmented is too high and the brightness difference between the targets to be segmented is large. Aiming at the image features under the Internet of things, a feature location method for variable threshold segmentation image based on improved wavelet is proposed. In this paper, a variable threshold algorithm is designed, which uses the multi-scale shrinkage threshold in the new spherical coordinate domain, and uses the adaptive nonlinear shrinkage function to continuously separate image information and noise information at the threshold. At last, the simulation experiment of this method is carried out, and a large number of comparisons with similar algorithms are made. The experimental results show that under the high noise image of the Internet of things, the improved image location method in this paper has better effect. The experimental results show that under severe occlusion and high noise conditions in the Internet of Things, the proposed method has better image feature location and denoising performance. When the noise intensity increases to 60%, the PSNR of the proposed method is 28.8764 dB. When the wavelet decomposition scale is 7, the average running time of the proposed method is 25 ms, and the denoising accuracy is 73%. It can effectively improve the peak signal-to-noise ratio and denoising accuracy, and shorten the running time.
All Science Journal Classification (ASJC) codes