In response to the problems of color distortion and blurred details in existing underwa-ter image enhancement methods,a convolutional network structure that utilizes multi-scale fea-tures and channel attention mechanism to enhance underwater images is proposed.The feature u-tilization is firstly improved by using features at different scales,and a channel attention mecha-nism is introduced to effectively improve the contrast restoration effect of the image in response to the different attenuation of the image three channels.Secondly,the network introduces residual connections and adds a random loss layer to avoid gradient vanishing.In order to prevent a single loss function from causing the model color correction to bias toward the background color,a joint loss function is designed.The experimental results show that the proposed method has shown im-provement in reducing detail loss and improving color cast.
维普
万方数据
