Threshold Based Text Region Identification and Extraction from Sign Board for Real Time Character Recognition
Neil Joshi, Neelesh Mehra, Vijay Kumar Yadav, S. B. Goyal · 2024
The text extraction and recognition from the image has very wide applications. For instance, it has been observed that in a new country or region having different native language, the language on sign boards is not understandable by the foreigners. So, if there will be a mechanism which capture the image of sign board and translate the extracted text written on sign board in the native language to desire language will be helpful. The image processing algorithms for text extraction found in literature are not suited for this application due to their computational complexity and cost. Also, they require costly hardware resources and large time. This paper presents a less complex and fast image processing algorithm for identification and extraction of the text area from image. The image captured in real time is pre-processed for noise removal. After pre- processing text are identified and extracted with the help of segmentation technique. The proposed algorithm uses global threshold segmentation method for text extraction process. The extracted text can be further processed for character recognition and speech to text conversion. The proposed algorithm is implemented in MATLAB software. The proposed state of art is less complex and suitable for handheld devices. The computation time for identification of text area and text extraction is approximately 6 to 8 second. This extracted text can be further used in many applications like text-tospeech for visually impaired individuals, in security and surveillance by taking id number from the allotted id cards, license plate recognition in automated parking, collecting data from medical images, and many more.