A fast algorithm for binary Segmentation using color information

We present the algorithm Segmentation by Maximization of Discriminant Function to perform the classification of a picture in two regions using color information. Overall classification algorithms require a training set, which underlies or sample for classification of objects. We develop an OpenCV application to provide small samples of an image using the computer mouse to create the training set; the rest of the image pixels will be classified using the given training set. Our goal is to perform classification in the shortest possible time with the least number of samples. So given a digital image, we want to separate one object in the image of the remaining elements. For example, in the case of a person with an undesirable background, separate the background to replace it by another. Our algorithm maximizes a linear function constrained by spatial coherence. Given the number of variables in the process image, segmentation is carried out using incremental images. We have achieved results in hundredths of a second. Results are reported using synthetic images to quantify the performance of the resulting image segmentation, and with two real images, which can be assessed visually.