【摘要】 基于暗通道先验原理,提出一种Shannon-Cosine小波结合精细积分法的农田图像去雾算法。针对现有透射率估计方法中存在的块效应以及复原后图像纹理丢失的问题,该算法对透射率图进行了细化,利用非线性偏微分方程保边特性,运用小波数值方法对其离散,降低方程组规模。并采用精细积分法求解,提高计算精度。达到了对透射率图局部平滑、边缘突出的目的。同时,对大气值A的计算方法进行改进,提高了运算速度。实验结果表明:本文算法得到的透射率图恢复的图像具有更好的清晰度,纹理更加丰富,相比于原暗通道算法,本文算法新增可见边之比提高了30. 36%,对比度提高了40. 72%,标准差提高了28. 21%,该算法可实现更好的去雾效果。更多还原

【Abstract】 Images collection of farmland is one of the important components of modern agricultural informatization. From the images,information such as the growth and distribution of crops and pests in the field can be monitored. Foggy weather is a special natural weather phenomenon. When the image of farmland is collected,fog is often caused,resulting in blurred and faded images. Aiming at this problem,based on the dark channel prior,a Shannon-Cosine wavelet precise integration method for farmland images dehazing was proposed. Aiming at the problems of the block effects in the transmission image and the loss of image texture after restoration,the transmission image was refined by the proposed algorithm.According to the characteristics of the transmission image,the nonlinear partial differential equation model was used to smooth and preserve the edge of images. The multi-scale Shannon-Cosine wavelet was used to discretize the equations. In this process,Shannon-Cosine wavelet can adaptively select feature points and identify the image texture to highlight the image texture features. This process reduced the size of the equations and the amount of computation. Then the precise integration method was used to solve the equations,and this method also effectively improved the calculation accuracy. The proposed algorithm also improved the atmospheric value A and improved the operation speed. The experimental results showed that the transmission image obtained by the algorithm had clear boundaries and was locally smooth. The recovered image had better definition and richer texture than the original algorithm.Compared with the original dark channel prior algorithm,the proposed algorithm increased the ratio of newly visible edges by 30. 36%,the contrast by 40. 72%,and the standard deviation by 28. 21%. The proposed algorithm had better dehazing results.更多还原