A New Threshold Selection Method Based on Fuzzy Expert Systems for Separating Text from the Background of Document Images

Threshold selection for text segmentation is an essential preprocessing step in most document processing algorithms. Although there are a number of sophisticatedly well-trained methods for text extraction, many researchers still prefer a fast and simple threshold selection algorithm for preprocessing. In this paper, a new global threshold selection method is proposed based on fuzzy expert systems (FESs). It initially enhances image contrast by using a FES. Then, the range of the threshold value is adjusted by using another FES and a pixel-counting algorithm. Finally, the threshold value is obtained as the middle value of the above range. To evaluate the performance of the proposed algorithm, we employed a database of 20 English, Farsi, and Chinese document images. Experimental results demonstrated that our method provided superior solution quality compared to a number of well-known frequently used counterpart algorithms. The computational burden of the proposed algorithms is actually light with the computational complexity of O(MN) for the image of size MxN.