In some implementations, faces based on image data from a camera of a mobile device are detected and one or more of the detected faces are determined to correspond to one or more people in a set of people that are classified as being important to a user. In response to determining that one or more of the detected faces correspond to one or more people in the set of people that are classified as being important to the user, quality scores are determined for the one or more detected faces that are determined to correspond to one or more people that are classified as important to the user. Multiple images with the camera are captured based on the quality scores such that, for each face determined to correspond to a person that is classified as important to the user, at least one of the multiple images includes an image of the face having at least a minimum quality score. A composite image is generated that combines the multiple images.