Improvements on Corner2, a lossless layout image compression algorithm for maskless lithography systems

Abstract

The Corner2 algorithm was designed to resolve the data delivery problem on maskless lithography systems. The Corner2 algorithm utilizes dictionary-based compression to handle repeated circuit components and applies a transform which is specifically tailored for layout images to deal with irregular circuit components. It obtains high compression ratios and fast encoding/decoding times while requiring limited decoder memory in the decoder hardware. Moreover, the entire decompression is simple so that it could be implemented as a hardware add-on to the lithography writer. However, there is some room for improvement in how we build the dictionary to handle frequent circuit patterns. In this paper, we introduce an improved way to discover frequent patterns from the circuit layout images based on binary integer programming. By applying this improved frequent pattern dictionary, we were able to obtain 4.5-35.8% more compression while maintaining the same Corner2 decoder. Moreover, this binary integer programming framework could be applied to other binary image compression problems with similar pattern restrictions.