Techniques are described for performing image processing on images of cameras located on or in a vehicle. An example technique includes receiving a first set of images obtained by a first camera and a second set of images obtained by a second camera; determining, for each image in the first set, a first set of features of a first object; determining, for each image in the second set, a second set of features of a second object; obtaining a third set of features of an object by combining the first set of features and the second set of features; obtaining a fourth set of features of the object by including one or more features of a light signal of the object; determining characteristic(s) indicated by the light signal; and causing a vehicle to perform a driving related operation based on the characteristic(s) of the object.