Determining pixels textual characteristics

摘要

Pixels are classified as text; or "non-text" by responding to accumulated luminous brightness gradients within a predetermined pixel interior. Text and non-text image characteristics rest on estimating text-containing rectangles by determining a narrow width and a height of a peak in an interval of accumulated gradients around each pixel. The text characteristics also are inferred by detecting character corners by quantifying and smoothing gradient angles and by associating directional corner responses with each pixel.