This contribution reports the EE1-1.1 test results of JVET-AM0131[1]. This contribution propose to leverage overlapped pixel features to enhance reconstruction quality while maintaining low complexity. By introducing innovative cropping and padding operations, the design captures spatial correlations at adjacent DCT block boundaries, addressing the limitation of traditional 2×2 DCT-based methods that ignore overlapping region information. The proposed model employs a grouped convolution fusion framework for Y and UV components, reducing computational redundancy and achieving a balance between complexity and performance.