夜间拍照图像质量增强算法的研究与实现

RESEARCH AND IMPLEMENTATION OF IMAGE ENHANCEMENT ALGORITHM FOR NIGHTTIME IMAGE

由于移动智能设备在夜间拍摄图像时,传统的闪光灯和无闪光灯条件下拍摄图像的融合算法在图像配准步骤中存在不准确以及在图像融合步骤中存在效果不佳的缺点。针对上述问题,提出一种夜间拍照图像质量增强算法。该算法首先提出基于ASIFT夜间图像配准算法来对闪光灯和无闪光灯条件下拍摄图像进行配准,避免传统配准方法的过拟合问题,然后用基于细节重构的图像融合算法对配准后图像进行融合,大大增强图像的细节信息。实验结果表明,基于ASIFT的夜间图像配准算法,可以提高夜间图像配准精确度,有效地避免出现重影等问题,而基于细节重构的图像融合算法有效地解决了色偏问题和细节提取问题。经过该算法处理后的图像既保留原来的光照氛围,又增强图像的细节信息,大大提高了图像的主客观质量。

When smart device take picture at night, the conventional image fusion algorithm of flash and no-flash image pairs has problems of inaccuracy of image registration and unsatisfied result of image enhancement. Thus, an image enhancement algorithm for nighttime image is proposed. Firstly, in order to avoid the over-fitting problem caused by feature points which are too concentrated, a nighttime image registration algorithm based on Affine-SIFT is proposed to register flash and no-flash image pairs. Next, the presented method used image fusion algorithm based on detail reconstruction to fuse the registered flash and no-flash image pairs. The experimental results show that the proposed algorithm can improve the accuracy of image registration and avoid the ghost in fused image. Furthermore, the proposed algorithm can improve the details in image and reduce color distortion. Flash and no-flash image pairs processed by the proposed algorithm can maintain ambient illumination in the no-flash image, improving the details and quality.