一种改进的渐进式图像超分辨率重建算法

An improved progressive image super-resolution reconstruction algorithm

针对单帧图像超分辨率重建不适定逆问题,提出一种改进的渐进式图像超分辨率重建算法。通过两层渐进式重建网络,第一层渐进是将整个重建网络分为放大重建和细节重建两个阶段,第二层渐进是在放大重建阶段采用两级放大策略对低分辨率图像进行粗放大。在放大重建阶段采用残差式网络,并在其残差块中加入注意力机制提升网络的性能。在细节重建阶段,先将网络提取的图像软边缘信息与第一阶段的粗重建结果相融合,再输入到细超分辨率重建网络中完成最终的图像重建。实验结果表明,与其他算法相比,改进算法对图像细节的重建效果好,在客观评价指标和视觉效果方面都有所提升。

In order to solve the highly ill-posed inverse problem in single image super-resolution(SISR)reconstruction,an improved two-layer progressive super-resolution reconstruction algorithm is proposed.The first progressive layer divides the whole reconstruction network into amplification reconstruction and detailed reconstruction.The second progressive layer roughly magnifies the low-resolution images in the amplification reconstruction stage by a two level magnification strategy.In this stage,the residual network is adopted,and an attention mechanism is added to improve the performance of the network.In the detail reconstruction stage,the image soft edge and the roughly reconstructed outputs of previous stage are fused and then fed into the image refinement network for high-quality SR image reconstruction.Experimental results show that compared with other algorithms,the proposed algorithm provides better image reconstruction results in terms of both objective evaluation and visual effects.