复杂背景下的图像文本区域定位方法研究

提出了一种基于YUV颜色空间与支持向量机的复杂背景文本区域定位方法。算法将图像由RGB颜色空间转换至YUV颜色空间；利用最小二乘法对图像的色调直方图曲线进行拟合并确定最佳拟合阶次，利用拟合后的曲线对图像进行颜色分层聚类；对分解出的各颜色图层分别进行处理，得到备选文本连通域；提取备选文本连通域的小波纹理特征并利用SVM进行文本判别。实验结果表明，提出的方法定位准确率在65%以上，可以有效地实现复杂背景下图像文本区域的定位。

Method for text region localization in complex background images

A method based on YUV color space and Support Vector Machine（SVM） is proposed for localizing text region in complex background images. The image is transformed from RGB color space to YUV color space. Curve fitting using least square method is carried out to fit the hue-level histogram of the image and then determine the best fitting order. In the next step, the image is classified into several layers including text layers and background layers. Each layer is processed respectively to pick up the potential text region. The wavelet texture feature is extracted from the candidate text region, and SVM is used to classify text and non-text regions. Experimental results show that the success rate of localization is above 65%. It can effectively localize the text region in complex background image.