Cellpose3 is a new method for image restoration that improves cellular segmentation. It addresses the challenge of segmenting degraded images, such as those with noise, blur, or low sampling, which are common in microscopy. Unlike previous methods that focus on restoring pixel values, Cellpose3 trains to output images that are well-segmented by generalist segmentation models while maintaining perceptual similarity to the original images. It is trained on a large, diverse dataset to ensure good generalization. Cellpose3 is implemented as "one-click" tools within the Cellpose graphical interface and API. The method uses a combination of segmentation and perceptual losses to train a denoising network. It first applies a denoising network to reduce noise, then uses a pre-trained Cellpose segmentation network to refine the output. This approach improves segmentation performance compared to traditional denoising methods like Noise2Self and Noise2Void. Cellpose3 also performs well in deblurring and upsampling tasks, outperforming other methods in both visual quality and segmentation accuracy. The method was tested on various datasets, including those with real Poisson noise, and showed significant improvements in segmentation performance. It was also evaluated on external data, demonstrating its effectiveness in improving segmentation of noisy images. Cellpose3 was found to be more effective than previous methods in denoising, deblurring, and upsampling, and it can be applied to a wide range of image types, including cells, nuclei, and tissues. The method was also tested on a variety of image restoration tasks, including deblurring and upsampling, and showed strong performance in these areas. It was found to be more effective than other methods in these tasks, and it can be used for a wide range of image types, including cells, nuclei, and tissues. Cellpose3 is a generalist method that can be used for a wide range of image restoration tasks, including denoising, deblurring, and upsampling. It is trained on a large, diverse dataset to ensure good generalization and is implemented as "one-click" tools within the Cellpose graphical interface and API. The method was found to be more effective than previous methods in denoising, deblurring, and upsampling, and it can be used for a wide range of image types, including cells, nuclei, and tissues.Cellpose3 is a new method for image restoration that improves cellular segmentation. It addresses the challenge of segmenting degraded images, such as those with noise, blur, or low sampling, which are common in microscopy. Unlike previous methods that focus on restoring pixel values, Cellpose3 trains to output images that are well-segmented by generalist segmentation models while maintaining perceptual similarity to the original images. It is trained on a large, diverse dataset to ensure good generalization. Cellpose3 is implemented as "one-click" tools within the Cellpose graphical interface and API. The method uses a combination of segmentation and perceptual losses to train a denoising network. It first applies a denoising network to reduce noise, then uses a pre-trained Cellpose segmentation network to refine the output. This approach improves segmentation performance compared to traditional denoising methods like Noise2Self and Noise2Void. Cellpose3 also performs well in deblurring and upsampling tasks, outperforming other methods in both visual quality and segmentation accuracy. The method was tested on various datasets, including those with real Poisson noise, and showed significant improvements in segmentation performance. It was also evaluated on external data, demonstrating its effectiveness in improving segmentation of noisy images. Cellpose3 was found to be more effective than previous methods in denoising, deblurring, and upsampling, and it can be applied to a wide range of image types, including cells, nuclei, and tissues. The method was also tested on a variety of image restoration tasks, including deblurring and upsampling, and showed strong performance in these areas. It was found to be more effective than other methods in these tasks, and it can be used for a wide range of image types, including cells, nuclei, and tissues. Cellpose3 is a generalist method that can be used for a wide range of image restoration tasks, including denoising, deblurring, and upsampling. It is trained on a large, diverse dataset to ensure good generalization and is implemented as "one-click" tools within the Cellpose graphical interface and API. The method was found to be more effective than previous methods in denoising, deblurring, and upsampling, and it can be used for a wide range of image types, including cells, nuclei, and tissues.