Generating a high-quality High Dynamic Range (HDR) image from several Low Dynamic Range (LDR) images is a challenging task due to the ghost artifacts caused by large motion and bad exposure among LDR images. To address this critical issue for ghost-free and high-quality HDR images, we propose an HDR Transformer Deformable Convolution (HDR-TDC) network which establishes the complicated alignment and fusion relationship between reference and non-reference images. Specifically, to solve the ghosting artifacts, we propose the Transformer Deformable Convolution Alignment Module (TDCAM) to extract the relevant content from the entire regions of non-reference images to align the multi-exposed frames. In addition, we propose the Spatial Attention Fusion Block (SAFB) to adaptively suppress the regions with bad exposure and select useful information across frames to effectively fuse multi-exposure features. Extensive experiments show that our method quantitatively and qualitatively achieves state-of-the-art performance.