一种加入标志识别的图像拼接技术

Image Mosaics Adding Logo Recognition

图像拼接过程中,由于两张图片曝光度和拍摄角度不同,对应匹配区域无法正确识别,进而导致图像融合失败或者融合时间过长。针对此问题,利用高斯滤波对图像平滑处理,起到减少图像噪声的作用。此时,处理后的图像就不会因为曝光度等拍摄问题导致特征点配准出现误差。此外,针对有明显相同特征区域的两副图像,提出要标志共有区域,并以此为中心通过距离和角度的关系进行迭代,去除基于尺度不变特征变换(SIFT)算法提取的两副图像不匹配的特征点,进而减少后续匹配时间和提高图像拼接成功率。实验结果表明该算法的可行性。

In image mosaic process, due to the different exposure and shooting angle of the two pictures, corresponding matching areas are unable to identify properly, then leading to image fusion failure or long fusion time. Aiming at this problem, Gaussian filter is used for the image smooth processing, playing the role of image noise reduction. At this point, the images after processing, because of shooting problems such as exposure, do not lead to feature points matching error. Besides, in view of two images which have the same characteristic region obviously. In this paper,signing a total area is proposed, iterating as a center through distance and angle, removing the mismatched feature points of the two images based on SIFT algorithm, then reducing subsequent match time and improving success rate of image mosaics. The experimental results show the feasibility of the algorithm.