Traditional lightweight image super-resolution reconstruction methods typically rely solely on single-scale convolutions to extract image features, simply aggregating shallow and deep features for image reconstruction. However, this approach overlooks the richness of receptive field information and the crucial role of intermediate latent features in image reconstruction, leading to limited interaction between convolutional kernels and consequently resulting in the loss of image detail information and low reconstruction accuracy. In light of this, this paper proposes a lightweight image super-resolution reconstruction method based on progressive receptive fields. The core of this method lies in the design of a staircase-like bidirectional convolutional chain structure, which gradually adjusts the size of the receptive field, effectively integrating the overall structural information and local detail features of the image, thus achieving diversified expression of information. Furthermore, a method for the fusion of multi-dimensional latent features is explored, aiming to fully exploit and utilize the correlations between multi-dimensional features. Experimental results demonstrate that compared to currently popular reconstruction methods, the proposed method excels in capturing image details. Particularly, under a scaling factor of 4, compared to the NGSwin, the proposed method requires fewer parameters, and on the Urban100 test set, PSNR and SSIM are improved by 0.09dB and 0.0027 respectively, further validating the superiority of the proposed approach.