基于特征点提取的图像增强算法

An Image Enhancement Algorithm for Feature Extraction

特征提取被广泛的运用在图像处理和计算机视觉领域。而特征点提取的效果很大程度上取决于图像的质量。在室外,考虑到多变的光照条件图像的质量通常很差。除此以外,物体间的互相遮挡也是特征提取上的一个难题。因此,研究怎样实时提取低对比度图像是一个很具挑战性的问题。Retinex理论被认为是分离光照图和反射物体图从而得到补充光照图的良好办法。但是亮度图像重建的计算复杂度很高。本文提出了一种基于Retinex理论的图像增强算法。新的算法优化了可变框架的Retinex理论。文中提出一种新的卷积函数,其能有效的减少计算时间,并将多尺度和卷积函数结合起来。文章的最后提出了一种能衡量提取特征点有效性的评估实验。实验结果证明,新的算法不仅能够极大的减少处理时间从而满足实时性要求,而且能使增强后的图像更适合与特征点提取。

Feature extraction has been widely used in the area of image processing and computer vision. And the effects of the feature extraction are greatly influenced by the quality of the image. In outdoor applications,the quality of image usually becomes poor due to the various illumination conditions. Besides illumination conditions,the overlap of objects is also a difficult for feature extraction. So it is attractive to study on feature extraction from low-contrast images,such as dark or overlap images adaptivel...