SExtractor is an automated software tool for source extraction from astronomical images, designed to detect, deblend, measure, and classify sources. It is particularly suited for large extragalactic surveys due to its ability to handle large images with minimal human intervention and a wide variety of object shapes and magnitudes. The software uses a combination of techniques, including background estimation, detection, deblending, filtering, photometry, and star/galaxy separation. Background estimation involves creating a "background map" to accurately detect faint objects and measure their fluxes. SExtractor uses a combination of κσ-clipping and mode estimation to construct this map, which is more robust than simple clipped means in crowded regions. Detection is performed using a one-pass algorithm to extract contiguous pixels from a template frame, which is created by convolving the original image with a suitable mask. This allows for the detection of both bright and low-surface-brightness objects. Deblending is used to separate merged objects, employing a multiple isophotal analysis technique to determine the real components of a composite object. This process involves re-thresholding the extracted pixels and deciding whether to split them based on their relative integrated intensity. Filtering is used to remove spurious detections by checking if an object would have been detected in the absence of neighbors. This is done by computing the contribution of each object to the mean surface brightness and subtracting it from the detection threshold. Photometry includes both isophotal and circular aperture magnitudes, as well as an adaptive aperture method and corrected isophotal magnitudes. These methods allow for accurate measurement of total magnitudes, with the adaptive aperture method being particularly effective for stars and galaxies. Star/galaxy separation is achieved using a neural network trained on simulated images, which allows for reliable classification of objects. The neural network is trained using a set of input parameters, including isophotal areas, peak intensity, and seeing information. The classifier is tested on both simulated and real images, showing high accuracy in separating stars from galaxies. SExtractor is efficient and robust, with a processing speed of about 40 kpix/sec on a SUN SPARC20 workstation. It is particularly well-suited for batch processing of large survey data due to its speed, robust deblending, and accurate estimation of total magnitudes. The software is available for use and its technical documentation is electronically accessible through the World Wide Web.SExtractor is an automated software tool for source extraction from astronomical images, designed to detect, deblend, measure, and classify sources. It is particularly suited for large extragalactic surveys due to its ability to handle large images with minimal human intervention and a wide variety of object shapes and magnitudes. The software uses a combination of techniques, including background estimation, detection, deblending, filtering, photometry, and star/galaxy separation. Background estimation involves creating a "background map" to accurately detect faint objects and measure their fluxes. SExtractor uses a combination of κσ-clipping and mode estimation to construct this map, which is more robust than simple clipped means in crowded regions. Detection is performed using a one-pass algorithm to extract contiguous pixels from a template frame, which is created by convolving the original image with a suitable mask. This allows for the detection of both bright and low-surface-brightness objects. Deblending is used to separate merged objects, employing a multiple isophotal analysis technique to determine the real components of a composite object. This process involves re-thresholding the extracted pixels and deciding whether to split them based on their relative integrated intensity. Filtering is used to remove spurious detections by checking if an object would have been detected in the absence of neighbors. This is done by computing the contribution of each object to the mean surface brightness and subtracting it from the detection threshold. Photometry includes both isophotal and circular aperture magnitudes, as well as an adaptive aperture method and corrected isophotal magnitudes. These methods allow for accurate measurement of total magnitudes, with the adaptive aperture method being particularly effective for stars and galaxies. Star/galaxy separation is achieved using a neural network trained on simulated images, which allows for reliable classification of objects. The neural network is trained using a set of input parameters, including isophotal areas, peak intensity, and seeing information. The classifier is tested on both simulated and real images, showing high accuracy in separating stars from galaxies. SExtractor is efficient and robust, with a processing speed of about 40 kpix/sec on a SUN SPARC20 workstation. It is particularly well-suited for batch processing of large survey data due to its speed, robust deblending, and accurate estimation of total magnitudes. The software is available for use and its technical documentation is electronically accessible through the World Wide Web.