Acquiring labels in anomaly detection tasks is expensive and challenging. Therefore, as an effective way to improve efficiency, pretraining is widely used in anomaly detection models， which enriches the model's representation capabilities, thereby enhancing both performance and efficiency in anomaly detection. In most pretraining methods, the decoder is typically randomly initialized. Drawing inspiration from the diffusion model, this paper proposed to use denoising as a task to pretrain the decoder in anomaly detection, which is trained to reconstruct the original noise-free input. Denoising requires the model to learn the structure, patterns, and related features of the data, particularly when training samples are limited. This paper explored two approaches on anomaly detection: simultaneous denoising pretraining for encoder and decoder, denoising pretraining for only decoder. Experimental results demonstrate the effectiveness of this method on improving model’s performance. Particularly, when the number of samples is limited, the improvement is more pronounced.

Due to the characteristics of color vegetation canopy images which have multiple details and Gaussion noise interference, the adaptive mean filtering (AMF) algorithm is used to perform the denoising experiments on noised images in RGB and YUV color space. Based on the single color characteristics of color vegetation canopy images, a simplified AMF algorithm is proposed in this paper to shorten the overall running time of the denoising algorithm by simplifying the adaptive denoising processing of the component V, which contains less image details. Experimental results show that this method can effectively reduce the running time of the algorithm while maintaining a good denoising effect.