基于边缘系数增强和对比度提升的Curvelet变换图像增强方法

Curvelet image enhancement based on edge coefficients enhancement and contrast improvement

对比度和边缘是二维图像的两个重要特性,为了提高图像的对比度和增强图像的边缘,本文提出了一种基于边缘系数增强和对比度提升的Curvelet变换图像增强方法。该方法首先对图像进行Curvelet变换得到低频系数和中高频系数,对低频系数采用自适应处理方法,使得低频系数按照原区间分布的概率重新调整分布区间,使低频系数的分布更加均匀,反变换后的图像对比度提高;对中高频系数先用传统的非线性增益函数进行增强处理,然后利用Curvelet系数的映射特性提取出和当前处理系数表达同一区域、同一方向的系数集合,计算这个集合的绝对值之和并与设定的阈值做比较,大于阈值则认为当前系数是边缘系数,之后对其进行进一步增强,这个处理方法能够提高边缘保护指数;最后进行Curvelet反变换得到增强后的图像。实验结果表明,与传统的Curvelet变换图像增强方法相比,运用本文方法增强的图像在主观观察上细节较清晰,边缘比较明显,客观指标中的对比度提升了20%左右,边缘保护指数提升了25%左右。

Contrast and edge of image are the two important properties of two-dimensional image. In order to improve the contrast and enhance the edge of the image, this paper proposed a method of Curvelet image enhancement based on edge coefficients enhancement and contrast improvement. The image was firstly transformed by Curvelet to obtain low frequency coefficients and medium-high frequency coefficients. The low frequency coefficients were processed by an adaptive processing method to readjust the interval distribution of the low frequency coefficients according to the probability distribution of original interval, producing a more uniform distribution of low frequency coefficients. And the image contrast was improved after the inverse transform. However, the medium-high frequency coefficients were enhanced by traditional nonlinear gain function. The set of coefficients which express the same area, the same direction with the current coefficient were extracted by the mapping properties of Curvelet coefficients. The absolute value of the set of coefficients was calculated and compared with the setting threshold, and the current coefficient was considered as edge coefficient if the value was greater than the threshold, and then, the current coefficient was further enhanced. This processing method could improve the edge-preserving index （EPI）. Finally, the enhanced image was obtained by Curvelet inverse transform. The experimental results showed that, compared with the traditional Curvelet transform image enhancement method, the method introduced in this paper provide more clear details and more obvious edges. The image contrast and EPI of objective indicators improved by about 20% and 25%, respectively.