A backside illuminated image sensor with an array of pixels formed in a substrate is provided. To improve surface planarity, bond pads formed at the periphery of the array of pixels may be recessed into a back surface of the substrate. The bond pads may be recessed into a semiconductor layer of the substrate, may be recessed into a window in the semiconductor layer, or may be recessed in a passivation layer and covered with non-conductive material such as resin. In order to further improve surface planarity, a window may be formed in the semiconductor layer at the periphery of the array of pixels, or scribe region, over alignment structures. By providing an image sensor with improved surface planarity, device yield and time-to-market may be improved, and window framing defects and microlens/color filter non-uniformity may be reduced.