Text recognition (OCR - Optical Character Recognition) is a research field that is gaining widespread attention due to its wide application in image and document processing. Although OCR technology has achieved a high level of success, the main challenge faced is the presence of noise in text image, noise causes decreased text recognition results, noise causes miss classification. Therefore needed noise handling text recognition.  The aim of this research is to provide valuable insight into the techniques and approaches used in the context of noise treatment using global threshold methods. The method used starts from an input digital image, then preprocessing is carried out by converting the image into a gray scale image, then a threshold is applied to the image, then recognition is carried out. From 6 experiments, the best results were obtained for character recognition with a threshold value (t) of 65 and a character recognition accuracy percentage of 94.29%. T value determined manually and static for separates the all object and the background, while in reality the lighting or contrast always varies. Suggestions for further research include developing an adaptive thresholding method approach to obtain threshold values automatically and optimally. So that if faced with varying lighting conditions or contrast, better results can be obtained.