A flat-top convex-bottom lower lens is formed by first applying a positive tone photoresist over a silicon oxide layer and an optional metallic barrier layer thereupon in a back-end-of-line (BEOL) metallization structure. The positive tone photoresist is exposed under defocused illumination conditions and/or employing a half-tone mask so that a cross-sectional profile of the positive tone photoresist after exposure contains a continuous and smooth concave profile, which is transferred into the underlying silicon oxide layer to form a concave cavity therein. After removing the photoresist, the cavity is filled with a high refractive index material such as silicon nitride, and planarized to form a flat-top convex-bottom lower lens. Various aluminum metal structures, a color filter, and a convex-top flat-bottom upper lens are thereafter formed so that the upper lens and the lower lens constitute a composite lens system.