Single Image Super-Resolution (SISR) made significant advancements in deep learning era, in which complex networks are designed consisting of a number of layers. Recently, a few researchers have started exploring Transformer for different vision tasks. However, it is proven that the transformer-based networks often rely on a limited spatial range of input data, suggesting the capabilities of Transformers are poorly harnessed in the existing models. To circumvent this limitation, we propose a transformer-based SR model that integrates channel attention alongside window-based self-attention. This combination leverages the global statistical insights with channel attention and the strong local feature modeling with window-based self-attention, exploiting their complementary advantages. Further, to improve the interaction between features across different windows, an overlapping cross-attention module is also utilized. Thus, it enhances the exchange of information between neighboring window features, leading to better feature aggregation. A consistent improvement over the existing methods was observed by conducting experiments with our proposed model on standard datasets. Moreover, with the proposed method, we secured 10th position in NTIRE 2024 Image Super Resolution ( \(\times 4\) ) Challenge [6].