Image analysis based on location sampling

作者: Daniel P. Huttenlocher , Peter C. Wayner

DOI:

关键词:

摘要: An image is analyzed by random or pseudorandom sampling. Information about attributes such as skew, typeface, and character size spacing obtained measuring a characteristic at number of sampled locations combining the results to obtain measure for image. The smaller than in image, but large enough achieve degree statistical significance. can be selected obtaining indicating candidate location then applying criterion determine whether an appropriate starting point measurement. For example, that must pixel within given color edge. measurement indicate distance from each directions, distances combined data distribution, averaging. divided into segments, which has prominent value being measured, first dividing sampling segments characteristic, grouping larger based on measurements. rectangles or, with words, include word paragraph.