Conventional source localization algorithms require a one-to-one source correspondence between films. This requirement makes source localization cumbersome and error prone because multiple sources must be carefully digitized and some sources can be obscured or missed. A new source localization algorithm is described in this paper. The algorithm fits a ribbon or needle image on film to a linear-quadratic equation, then analytically determines the 3-D ribbon locus by its image on the other projection, and finally localizes the sources in the ribbon by tracing along the ribbon image. Only three points per ribbon per film are required, and corresponding points need not be identified on the other film. Phantom experiments and tests on clinical cases demonstrate that the source localization algorithm can increase the efficiency by a factor of up to 5, improve accuracy to about 1 mm, and reconstruct obscured or shifted sources without decreased accuracy and efficiency. The simplicity and minimal entry of data make this technique desirable for clinical use.