Simulations of Single Molecule Localization Microscopy frames with scattered single emitters

<p>Datasets used in the work &quot;Combining deep learning with SUPPOSe and Compressed Sensing for SNR-enhanced localization of overlapping emitters&quot;.</p> <p>The file <strong>Sample dataset.zip</strong> contains simulated images of frames of Single-Molecule Localization Microscopy. In the directory structure, Q is the number of emitters, d is the emitter distance (in pixels), imax is the image maximum intensity and i indexes different noise realizations. There are three&nbsp;images within each folder: <strong>X</strong> is a noiseless image, <strong>Y</strong> is the image with noise and background and <strong>Z</strong> is a denoised image predicted using a convolutional neural network.</p> <p>The file <strong>Train dataset.zip</strong> contains 5000 simulated pairs of images of single emitters distributed randomly that were used&nbsp;to train a convolutional neural network for denoising. The folder <strong>X</strong> contains&nbsp;noiseless images and the folder&nbsp;<strong>Y</strong>&nbsp;contains the corresponding images with noise and background.</p> <p>In all cases, a Gaussian PSF with size&nbsp;<span class="math-tex">\(\sigma = 3\)</span> px was used as the PSF of the imaging system, Noise is modeled as a Poisson process with&nbsp;a dark signal <span class="math-tex">\(i_{dark} = 10\)</span>.</p> <p>Corresponding author: Axel M. Lacapmesure (alacapmesure@fi.uba.ar)</p> <p>&nbsp;</p> <p><strong>CHANGELIST</strong></p> <ul> <li>Version 2: corrected all file extensions in &quot;Train dataset.zip&quot; that were wrong.</li> </ul>