An automatic visual inspection system for integrated circuit chips

Abstract

A formulation of an automatic visual inspection and final packaging system is presented in this paper. An integrated system on the basis of such a formulation is then proposed for the automatic visual inspection and wire-bonding of integrated circuit (IC) chips. An input chip image is first segmented so that the IC image can be extracted. The image registration is then carried out in two steps. The macroregistration is achieved by locating the mask frame first, and each mask subpattern is microregistered independently by means of the proposed unnormalized cross correlation of contrast method. A task-dependent, context-dependent image filter is designed to transform the IC image into another image so that at any moment only relevant information is emphasized while irrelevant information is suppressed. Such a transformed or coded image is then examined to detect existent IC defects. Each IC chip under inspection is then classified into one of three categories: accept, or reject, or to be reworked. Finally, each accepted IC chip is soldered to a package substrate for wire-bonding.