We present some new ideas on pattern recognition of the layered track chambers. Our method is a modification of that of global transformations and makes valid use of not only positional, but also directional information of track segments. We apply this method to the VENUS drift chamber at TRISTAN and demonstrate that it is fast and powerful for pattern recognition of high multiplicity events with complicated e +e - jets. The reconstruction efficiency obtained is 93% for hadron events simulated by Monte Carlo program and the processing time obtained is 12 ms per track using a FACOM M382 computer. This approach is also examined using cosmic ray events.