小波变换域井下视频监控图像改进阈值去噪方法

发布时间：2018-11-20 07:08

【摘要】：通过对井下视频监控系统实时采集的各类开采、地质环境信息进行有效分析,可为制定井下开采方案以及进行灾害救援提供准确依据。由于井下光照不均匀,粉尘较多,严重干扰了监控探头获取高质量的监控图像。为此,结合小波变换方法,提出了一种小波变换域改进阈值去噪方法。该方法首先对井下视频监控图像进行3层小波分解,得到低频和高频分解系数;其次对低频分解系数进行重构,得到空间域背景图像,采用维纳滤波算法进行处理,以去除其中存在的少量噪声;然后根据经典小波硬、软阈值去噪模型的不足,提出了一种改进型阈值去噪模型,该模型可分别根据不同的高频分解系数自适应设定阈值,可有效去除不同分解层高频系数中的噪声,对去噪后的各高频分解系数进行重构,得到空间域细节图像;最后,分别将去噪后的空间域背景图像和细节图像进行叠加,得到去噪后的井下视频监控图像。采用1幅内蒙古某煤矿井下视频监控图像进行试验,并引入了经典小波硬、软阈值去噪模型及2类已有的改进型阈值去噪模型与所提方法进行试验对比,结果表明,所提方法不仅可有效去除井下视频监控图像中的噪声,而且可保持图像细节信息完整性。
[Abstract]:Through the effective analysis of all kinds of mining and geological environment information collected in real time by the video surveillance system, it can provide an accurate basis for drawing up the downhole mining plan and carrying out disaster rescue. Due to uneven lighting and more dust underground, the monitoring probe is seriously disturbed to obtain high quality monitoring images. In this paper, an improved threshold denoising method in wavelet transform domain is proposed. The method firstly decomposes the video surveillance image into three layers of wavelet transform to get the decomposition coefficients of low frequency and high frequency. Secondly, the low frequency decomposition coefficient is reconstructed, and the spatial background image is obtained, and the Wiener filter algorithm is used to remove a small amount of noise. Then, according to the shortcomings of the classical wavelet hard and soft threshold denoising model, an improved threshold de-noising model is proposed, which can adaptively set the threshold according to different high frequency decomposition coefficients. It can effectively remove the noise from the high frequency coefficients of different decomposition layers and reconstruct each high frequency decomposition coefficient after denoising to obtain the spatial domain detail image. Finally, the background image and detail image after denoising are superposed to obtain the downhole video surveillance image after denoising. A video surveillance image of a coal mine in Inner Mongolia is tested, and the classical wavelet hard, soft threshold denoising model and two kinds of improved threshold denoising model are introduced and compared with the proposed method. The proposed method can not only effectively remove noise from underground video surveillance images, but also maintain the integrity of image detail information.
【作者单位】：吉林省经济管理干部学院计算机实验中心;
【分类号】：TD76

【参考文献】