A method for forming an FSI image sensor device structure is provided. The method includes forming a pixel region in a substrate and forming a dielectric layer over the substrate. The method includes forming a trench through the dielectric layer, and the trench includes a top portion and a bottom portion, and the trench is directly above the pixel region. The method includes forming a protection layer in the bottom portion of the trench and enlarging a top width of the top portion of the trench, and the trench has a wide top portion and a narrow bottom portion. The wide top portion has top sidewall surfaces, the narrow bottom portion has bottom sidewall surfaces, and the top sidewall surfaces taper gradually toward the bottom sidewall surfaces. The method includes filling a transparent dielectric layer in the trench to form a light pipe.