Construction method of improved generative adversarial network for image super-resolution reconstruction

• Main Authors: ZHANG JIAHAO, XIANG QIANG, ZENG NINGZHI, WEN JING, HU KUN, LU JINZHENG, WEI LIJUAN, WANG ZIKANG
• Format: Patent
• Language: Chinese; English
• Published: 19.07.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The invention discloses a construction method of an improved generative adversarial network for image super-resolution reconstruction, and the method comprises the steps: obtaining a plurality of target images, and constructing an image data set based on the plurality of target images; an SRGAN network used for image super-resolution reconstruction is constructed; introducing a self-attention mechanism into the residual dense module to obtain an attention enhanced residual dense module; and modifying a feature extraction part in the SRGAN network by adopting an attention enhancement residual dense module, and constructing an attention enhancement residual dense generative adversarial network AEDRGAN. According to the invention, a self-attention mechanism is introduced into the residual module, the attention relationship is utilized, the dependency relationship and importance among pixels are captured, and the dense residual network is introduced into the generator, so that the image feature taking is increase