Still today, bitonal images (black & white images without any shades in between - just black and white) of high quality are the precondition for trouble-free automatic text recognition (OCR). The challenge is to distinguish between relevant information and background information - at the same time, however, image processing may not create damage to character structures. Due to the high calculation amount and the necessity to use uncompressed images as source image for processing, mainly hardware-oriented image processing methods were used in the past, and they are widely in use still today. Due to the risen PC calculation power and by optimizing calculation steps, a method could be developed for dynamic thresholding of grayscale images that returns optimal results even from extremely difficult images; in the spring of 2003, it delivered about 600 images per minute (A4,200dpi). © Springer-Verlag Berlin Heidelberg 2004.
CITATION STYLE
Woitha, D., & Janich, D. (2004). Adaptive threshold. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2956, 14–25. https://doi.org/10.1007/978-3-540-24642-8_2
Mendeley helps you to discover research relevant for your work.