Optical Character Recognition (OCR) technology is commonly used to convert printed files to digital; format, however this technology still has the problem of gaining high accuracy especially on complex document; elements such as Arabic posters. This paper introduces a new approach, which synergistically integrates three powerful techniques: We implement state-of-the-art segmentation, the YOLO object detector, and the Tesseract OCR engine. The resulting methodology is a complex multistage pipeline, which first applies a fine-tuned YOLO model, trained on a large dataset of Arabic posters, for very accurate detection and localization of the text areas present in more or less sophisticated poster layouts. And after detecting the text we implement some novel segmentation methods to separate the actual words from its complex background to make them recognize by the Tesseract engine. Our method on a large set of various Arabic posters and apply Tesseract OCR alone and with Live Scribe on the same set of posters. The outcome is remarkable: 84.6% recognition rate when compared to 64.8% recognition rate for only Tesseract. Nevertheless, this major gain is realized at competitive processing times, making our approach relevant and feasible for usage addressing real-world problems. Moreover, we compare our method with both open sourced and commercial OCR systems for Arabic poster layouts and prove its superiority in overcoming the specific challenges posed by Arabic posters, and, to our knowledge, achieving the new optimal performance in accuracy.