基于多通道视觉注意力的图像清晰度复原方法

Image Clear Recovery Method Based on Multi-Channel Visual Attention

图像采集过程中被拍摄物和照相机之间的相对运动会造成图像的模糊,严重影响图像清晰度。由于图像视觉降质后特征较为模糊,其复原难度较大。多通道模式可以将含有通道的图像分割成单个的通道,有效降低图像复原复杂度。为此,提出基于多通道视觉注意力的图像清晰度复原方法。根据视觉注意力下多通道分配函数,生成图像特征视觉信号和信息权重。将图像视觉状态传入长短期记忆网络,随机生成视觉信息元素。基于此,将注意力模型多通道化,弥补图像视角缺失,计算图像自然光向量及透射率。为使去雾后的图像自然清晰,加入调节参数调整透射率,利用给定灰度方差计算图像透射率平滑程度,通过高斯函数卷积获得最终清晰度复原图像。经实验证明,所提方法的图像复原效果更好,去雾明显,图像复原后透射率高,图像细节信息增加。

In the process of image acquisition,the relative motion between subject and camera may causes image blur problem,seriously affecting the image definition.However,it is difficult to restore the degraded image.At pres-ent,the multi-channel mode can segment the image containing channels into several single channels,reducing the complexity of image restoration.In this paper,a method of restoring image clarity based on multi-channel visual at-tention was proposed.According to assignment functions under multi-channel visual attention,the visual signal and information weight of image feature were generated.Then,the visual state of image was transferred into the long-term and short-term memory network,so that the visual information elements can be generated randomly.On this basis,multiple channels were introduced into the attention model to make up for the lack of visual angle,thus calculating the light vector and transmittance of the image.In order to make the image more natural and clearer after defogging,the adjustment parameters were added to adjust the transmittance.Finally,the smoothness of the image transmittance was calculated by using the given gray variance,and then the restored image was obtained by Gaussian function con-volution.Experiment results prove that the proposed method has better image restoration,obvious fog removal and high transmittance,increasing image details after image restoration.