一种自适应非线性彩色图像增强技术

An Adaptive and Nonlinear Technique for Color Image Enhancement

在图像处理和模式识别中,经常会遇到图像有过亮或过暗的区域。不仅会使图像的视觉效果不佳,而且会造成图像特征的丢失。提出一种新的自适应非线性彩色图像增强算法。新算法包括五个步骤:将彩色图像转换为灰度图像;利用非线性函数对图像增强;利用高斯核函数与图像进行卷积得到图像的邻域信息;根据邻域信息对图像进行自适应增强;图像色彩还原。最后,利用了基于SNoW的人脸定位器对算法的效果进行了验证。实验表明这种图像增强技术能够提高人脸定位器的准确率。

In image processing and pattern recognition, extremely bright or dark regions may exist in images, which will result in not only the degradation of visual effect but also loss of some features of the images. This paper presents a novel color image enhancement algorithm based on an adaptive and nonlinear technique, which consists of five stages： converting color images to grayscale images; intensity transforming based on a specifically designed nonlinear transfer function ; obtaining intensity information of neighborhood by a convolution process with a Gaussian kernel; enhancing contrast according to the information of neighborhood; restoring the color by a linear function. Finally, this algorithm is tested by SNoW - based face detector. The experiments show that the application of this image enhancement algorithm can effectively improve the hit rate of the face detection.