Aiming at the deficiency of the traditional NLM algorithm that the redundant information of image structure is not fully utilized

an improved NLM denoising algorithm based on linear prediction is proposed. According to the characteristics of image spatial structure

the image signal is transformed from two-dimensional to one-dimensional by column

and the visual features are extracted by linear prediction

so as to make full use of the redundant information of image structure and optimize the denoising characteristics of NLM algorithm. The results show that: when the noise intensity is large

the denoising effect of the improved NLM algorithm is close to the optimal performance of the image; The improved NLM algorithm depends on the order of linear prediction

When the order increases in a certain range

the denoising effect is better. In view of the high-intensity noise in the complex environment

the image can achieve ideal noise reduction effect.