无人机航拍图像超分辨率重建算法研究
|
摘要
针对影响无人机航拍图像质量的主要影响因素,提出改进的去雾算法和改进的快速超分辨率重建算法相结合的快速无人机航拍图像超分辨率算法。首先,设计分类器,分辨有雾和无雾图像,并利用暗通道先验,分割天空区域并优化透射率进行去雾,同时进行对比度校正,取得了较好的去雾效果,克服了色偏、天空区域透射率估计不准确等问题。然后,设计快速超分辨率重建神经网络,以原始低分辨率去雾图像作为输入,经过特征提取、收缩、特征映射、亚像素卷积重组等处理,实现了快速超分辨率重建,克服了同类算法参数多、运算量大、训练速度慢等问题。实验结果对比表明算法的精度超过了同类算法,运算速度也接近同类算法最好水平。
According the main degradation factors of UAV image,proposed a fast super-resolution algorithm for UAV images,which combines the improved defogging algorithm and improved fast superresolution reconstruction algorithm. Firstly,dark channel priori is used to segment the sky region and optimize the transmittance for defogging,and correct the contrast of result,which achieves better defogging effect and overcomes the problems of color deviation and inaccurate estimation of transmittance in sky region. Then,a fast super-resolution reconstruction neural net-work is designed,which takes the original low-resolution defogging image as input. After feature extraction,shrinkage,feature mapping,sub-pixel convolution and reconstruction, the fast super-resolution reconstruction algorithm is implemented. The proposed algorithm overcomes the problems of many algorithms,such as large parameters,large computation,and slow training. Through the comparison of experimental results,the accuracy of the algorithm exceeds the similar algorithms,and the computing speed is close to the best of the similar algorithms.
According the main degradation factors of UAV image,proposed a fast super-resolution algorithm for UAV images,which combines the improved defogging algorithm and improved fast superresolution reconstruction algorithm. Firstly,dark channel priori is used to segment the sky region and optimize the transmittance for defogging,and correct the contrast of result,which achieves better defogging effect and overcomes the problems of color deviation and inaccurate estimation of transmittance in sky region. Then,a fast super-resolution reconstruction neural net-work is designed,which takes the original low-resolution defogging image as input. After feature extraction,shrinkage,feature mapping,sub-pixel convolution and reconstruction, the fast super-resolution reconstruction algorithm is implemented. The proposed algorithm overcomes the problems of many algorithms,such as large parameters,large computation,and slow training. Through the comparison of experimental results,the accuracy of the algorithm exceeds the similar algorithms,and the computing speed is close to the best of the similar algorithms.
引文
[1] He K,Sun J,Tang X. Single image haze removal using dark channel prior[J]. IEEE Trans Pattern Anal Mach Intell,2011,33(12):2341-2353.
[2] Xu J,Zhao W,Liu P,et al. Removing rain and snow in a single image using guided filter[C]//IEEE International Conference on Computer Science and Automation Engineering,2012:304-307.
[3]吴迪,朱青松.图像去雾的最新研究进展[J].自动化学报,2015,41(2):221-239.
[4] Narasimhan,Srinivasa G,Nayar,et al. Vision and the atmosphere[J]. International Journal of Computer Vision,2002,48(3):233-254.
[5] He K,Sun J,Tang X. Guided image filtering[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(6):1397-1409.
[6] Dong C,Loy C C,He K,et al. Image superresolution using deep convolutional networks[J].IEEE Trans Pattern Anal Mach Intell,2016,38(2):295-307.
[7]肖进胜,高威,邹白昱,等.基于天空约束暗通道先验的图像去雾[J].电子学报,2017,45(2):346-352.
[8]彭国福,林正浩.图像处理中Gamma校正的研究和实现[J].信息化研究,2006,32(2):30-32.
[9] Dong C,Chen C L,Tang X. Accelerating the super-resolution convolutional neural network[C].European Conference on Computer Vision,2016:391-407.
[10]Lim B,Son S,Kim H,et al. Enhanced deep residual networks for single image super-resolution[C].Computer Vision and Pattern Recognition Workshops,2017:1132-1140.
[11]Shi W,Caballero J,Huszar F,et al. Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network[C]. Computer Vision and Pattern Recognition,2016:1874-1883.
[12]Long J,Shelhamer E,Darrell T. Fully convolutional networks for semantic segmentation[C]. IEEE Conference on Computer Vision and Pattern Recognition,2015:3431-3440.
[13]Xu J,Zhao Y,Dong Y,etal. Fast and Accurate Image Super-Resolution Using a Combined Loss[C]. Computer Vision and Pattern Recognition Workshops,2017:1093-1099.
[14]Yang J,Wright J,Huang T S,Ma Y. Image SuperResolution Via Sparse Representation[J]. IEEE Transactions on Image Processing,2010,19(11):2861-2873.
[15]Tan Z,Bai X,Wang B,et al Fast Single-image Defogging[J].Fujitsu Scientific&Technical Journal,2014,50(1):60-65.
[16]De Boor C. Bicubic spline interpolation[J]. J.math.physics,1962,41(3):212–218.
[17]王国珲,谢倩.近点光源照射下球面坐标PSFS三维形貌重建方法[J].西安工业大学学报,2018(1):14-19,34.
[18]赵文仓,王倩,董瑞霞.基于多尺度分析的人脸识别算法研究[J].工业仪表与自动化装置,2017(4):11-15.
[2] Xu J,Zhao W,Liu P,et al. Removing rain and snow in a single image using guided filter[C]//IEEE International Conference on Computer Science and Automation Engineering,2012:304-307.
[3]吴迪,朱青松.图像去雾的最新研究进展[J].自动化学报,2015,41(2):221-239.
[4] Narasimhan,Srinivasa G,Nayar,et al. Vision and the atmosphere[J]. International Journal of Computer Vision,2002,48(3):233-254.
[5] He K,Sun J,Tang X. Guided image filtering[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(6):1397-1409.
[6] Dong C,Loy C C,He K,et al. Image superresolution using deep convolutional networks[J].IEEE Trans Pattern Anal Mach Intell,2016,38(2):295-307.
[7]肖进胜,高威,邹白昱,等.基于天空约束暗通道先验的图像去雾[J].电子学报,2017,45(2):346-352.
[8]彭国福,林正浩.图像处理中Gamma校正的研究和实现[J].信息化研究,2006,32(2):30-32.
[9] Dong C,Chen C L,Tang X. Accelerating the super-resolution convolutional neural network[C].European Conference on Computer Vision,2016:391-407.
[10]Lim B,Son S,Kim H,et al. Enhanced deep residual networks for single image super-resolution[C].Computer Vision and Pattern Recognition Workshops,2017:1132-1140.
[11]Shi W,Caballero J,Huszar F,et al. Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network[C]. Computer Vision and Pattern Recognition,2016:1874-1883.
[12]Long J,Shelhamer E,Darrell T. Fully convolutional networks for semantic segmentation[C]. IEEE Conference on Computer Vision and Pattern Recognition,2015:3431-3440.
[13]Xu J,Zhao Y,Dong Y,etal. Fast and Accurate Image Super-Resolution Using a Combined Loss[C]. Computer Vision and Pattern Recognition Workshops,2017:1093-1099.
[14]Yang J,Wright J,Huang T S,Ma Y. Image SuperResolution Via Sparse Representation[J]. IEEE Transactions on Image Processing,2010,19(11):2861-2873.
[15]Tan Z,Bai X,Wang B,et al Fast Single-image Defogging[J].Fujitsu Scientific&Technical Journal,2014,50(1):60-65.
[16]De Boor C. Bicubic spline interpolation[J]. J.math.physics,1962,41(3):212–218.
[17]王国珲,谢倩.近点光源照射下球面坐标PSFS三维形貌重建方法[J].西安工业大学学报,2018(1):14-19,34.
[18]赵文仓,王倩,董瑞霞.基于多尺度分析的人脸识别算法研究[J].工业仪表与自动化装置,2017(4):11-15.