To address the challenges of object detection in complex artistic images such as Thangka, including significant directional textures, large scale variations, and similar fine-grained features, this study proposes a global-directional hybrid attention module named Pamamba. Built upon the linear-complexity Mamba architecture, the module consists of the enhanced state space module (ESSM) and channel attention feedforward network (CAFFN), forming a spatial-channel collaborative enhancement mechanism. ESSM adopts a dual-path architecture: the global path employs the deep feature extraction module (DFEM) to capture global contextual semantics, while the directional enhancement path integrates the directional information enhancement module (DIEM) with a diagonal scanning mechanism (E-SS2D) to compensate for the directional perception limitations of conventional Mamba. CAFFN introduces a gating mechanism and dilated convolutions to dynamically enhance channel-wise feature interaction and multi-scale representation. Pamamba is a plug-and-play module that can be embedded into mainstream detectors such as YOLOv8/v10/11. Experimental results on the COCO benchmark and a self-built Thangka dataset demonstrate that the proposed method significantly improves detection performance and achieves mAP@0.5 of up to 66.9% and mAP@[0.5:0.95] of up to 60.6% on the Thangka dataset, especially improving small-object recall. It provides an effective tool for robust object detection in complex scenarios.