Cheng Liu

TL;DR: A weighted edge-based level set method based on multi-local statistical information to better segment noisy images and provides higher segmentation accuracies and more accurate segmentation results, which demonstrate its effectiveness and robustness.

TL;DR: In this article , an improved mixed convolution is introduced into the IRes (inverted residual block) to form an MIRes (mixed inverted residual block), which is used to replace the Res (residual block) in the deep CSP module of the backbone network to enhance the multiscale feature extraction capability of backbone network.

TL;DR: In this article, an improved edge lever set-based method and system for segmenting a noisy image is presented, which includes the following steps: pretreatment of the noisy image, initialization of a level set function, calculation of a local region fitting mean value; calculation of an edge-stopping function; update of the level-set function; determination of level set evolution termination condition and output of a segmentation result.

TL;DR: This paper adopts groupnormalization (GN) instead of batch normalization (BN) in the UNet network to alleviate the impact of the model on batch size and results show that the improved model (UNet-GN) improves the mean intersection over union (MIoU) and mean pixel accuracy (MPA) compared with the original model.