The research on real-word super-resolution further promotes the application of super-resolution and becomes a hotspot. A real-world super-resolution method based on residual attention network with tree based multi-branch structure is proposed to solve the problem that a single output is difficult to ensure stable and accurate high-frequency details. Tree based structure is designed to form a multi-branch super-resolution network, which can enhance feature representation capability and enrich high-frequency details of the restored image. A dual residual path schema is proposed for basic blocks, which lets more low frequencies pass. The basic block adopts a dense residual module and attention mechanism to deepen the network and to adaptively adjust global information in both channel and space. Experimental results show that the proposed method can achieve better reconstruction than the current advanced similar methods.