Abstract: Due to the problem of high complexity and lack of adaptability, an algorithm based on adaptive LBP for video text detection is put forward. The algorithm uses the divided difference of global and local pixel grey to determine the size of adaptive threshold. However, it, having strong adaptability, can maximize the removal of background. At the same time, to reduce the complexity, approximate round masking which simplifies the algorithm by covering multi-direction judgments is presented. Experimental results show that the proposed algorithm can detect characters accurately even in video with complex background, and it is of strong adaptability and good practicality, also has low complexity.

Key words: character detection, LBP operator, adaptive threshold, seed pixel, geometrical properties

摘要： 针对LBP算法自适应性弱及复杂度高的问题，提出一种基于自适应LBP算子的视频文本检测算法。该算法利用全局及局部的像素灰度均差决定自适应阈值大小，能最大限度去除复杂背景，自适应性较强。给出基于近似圆的掩模算法来覆盖多方向种子生长区，降低其复杂度。实验结果表明，该算法在复杂背景下能取得较高的检测率，具有良好的自适应性和实用性，复杂度较低。

关键词: 字符检测, LBP算子, 适应阈值, 种子像素, 几何特性

CLC Number: 

• TP18