Abstract: When the denoising algorithm based on principal component analysis (PCA) performs local pixel grouping (LPG), due to noise uncertainty and randomness,taking Euclidean distance directly as judgment standard of image block similarity is easy to make the results bias. To solve this problem, an LPG-PCA image denoising algorithm based on vector similarity is proposed. The combination of vector similarity and Euclidean distance is used as the judgment standard of similar image blocks, and the selection of similar image blocks is optimized. In addition, the adaptive number selection method is adopted to select the sample number of similar image blocks, which makes the selection of sample number more reasonable and further improves the quality of image denoising. Experimental results show that the proposed algorithm is superior to the traditional LPG-PCA image denoising algorithm in terms of peak signal-to-noise ratio and structural similarity, and also has a certain denoising effect on sub millimeter wave imaging.