This study proposes a novel method for semantic line correction utilizing deep Hough transform, aimed at tackling the challenges associated with detecting semantic straight lines and correcting image distortions in natural scenes. Traditional approaches frequently consider semantic straight line detection as a subset of object detection or simply adapt conventional object detection techniques, thereby neglecting the inherent characteristics of straight lines and consequently leading to suboptimal performance. Herein, we employ a deep Hough transform-based algorithm to achieve semantic line detection in images. The adopted approach utilizes parameterization and the Hough transform to map depth representations into parameter space for straight line detection, effectively exploiting the geometric properties of lines. Innovatively, we introduce the Distortion Correction Sub-network (DTN) to mitigate image distortion and enhance the success rate of deep Hough transform line detection. Furthermore, the DTN can dynamically adjust its spatial transformation according to various image transformations, thereby achieving effective image distortion correction. Experimental results demonstrate that the proposed method outperforms previous state-of-the-art methods on both self-constructed and publicly available datasets, thus substantiating its efficacy and superiority in addressing the challenges of semantic line detection and image distortion correction.