基于分层随机Hough变换的文本图像倾斜检测

传统的Hough变换在文本图像倾斜检测时存在计算量大、速度慢的问题,不适用于实际中的批量处理。本文提出一种基于分层的随机Hough变换文本图像倾斜检测方法,首先运用变分辨率金字塔策略,把图像分解成具有不同分辨率的多层子图像,然后对每层图像进行随机Hough变换,并采用两点法表示直线使参数空间降到一维,并通过降分辨率图像的快速检测结果,设置合理的检测门限以减少无效累积。实验结果表明,改进后的方法在保证检测精度相近的前提下,其计算速度比传统方法快十倍左右。

Skew detection of document image based on multi-level RHT

Traditional Hough transform has been limited by calculations,therefore it is too slow for the skew detection of a document image.It is not suitable for real time batch processing.A new skew detection method of document image based on multi-level randomized Hough transform is proposed in this paper.The method employs multi-resolution image pyramid at first,and then makes random Hough transform on each level image of pyramid.We expressed a line using two-point method so that to reduce the parameter space from two-dimensions into one-dimension,and then decrease inefficient accumulation by using a reasonable voting threshold.The experimental results show that the improved method is not only faster than traditional method,but also can ensure detection accuracy.