You can use the face detection algorithms from Computer Vision Toolbox to detect the face. There are a number of of examples provided in that toolbox.
It is common that individual cropped images from webcam streams are too low of resolution to be used for ID card purposes.
You have several options:
1) use a higher resolution webcam
2) use a stream of images with either the camera or the person moving a bit in order to reconstruct a higher resolution image. Although at any one time the output quality is limited by the resolution of the sensor, if you can move the sensor relative to the image by a fraction of the width of a pixel then you can combine the information to bring out details;
3) you can take a single image and apply Deep Learning techniques to it in order to predict what the person might look like in higher resolution. You might be to construct an image that looks good, but such images should never be used for official ID purposes, because they do not accurately represent the face of the actual person.
