Underwater image enhancement method based on residual dual-attention network

• Main Authors: ZHAO MINGHUA, SHI ZHENGHAO, SHANG RONGRONG, YOU ZHENZHEN, FENG YANING, GE FEIHANG
• Format: Patent
• Language: Chinese; English
• Published: 27.08.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention discloses an underwater image enhancement method based on a residual dual-attention network. The underwater image enhancement method is specifically implemented according to the following steps: step 1, preprocessing an input image; 2, mapping the pre-processed image to an R channel, a G channel and a B channel, and respectively carrying out independent convolution operation on each color channel; step 3, processing the image by using a deep network architecture of a residual double-attention-based codec, the encoder being used for extracting image features, reducing spatial dimensions and increasing a receptive field, wherein the decoder is used for gradually recovering detail information and spatial dimensions of the recovered image; training a residual dual-attention network model by using the preprocessed image to obtain a trained residual dual-attention network model; and step 4, inputting a fuzzy low-quality underwater image into the trained residual dual-attention network model to obtain