An Adaptive Noise Reduction Method for Depth Data Based on the Pulse-Coupled Neural Network
Xiang Li, Weijie Wang · 2024
The paper proposed a noise reduction algorithm for depth data affected by noise, leveraging the Pulse-Coupled Neural Network in conjunction with an adaptive weighted median filtering approach. This method initiates the process by identifying noise points within the depth data using PCNN. Subsequently, it determines the coordinates of the filter window based on the locations of these noise points. Varied weights are then assigned in accordance with the quantity of noise data present within the filtering window. Ultimately, the weighted median filtering method is employed to process the noise data within the window, facilitating adaptive noise reduction. Experimental results illustrate that this innovative noise reduction algorithm outperforms the classic median filter algorithm across a range of noise intensities. Furthermore, it demonstrates exceptional noise reduction performance while demanding minimal computational resources.