To determine the optimum parameters for extracting three carotenoids including zeaxanthin, lutein epoxide, and violaxanthin from pepper leaves by response surface methodology (RSM), a solvent of acetone and ethyl acetate (1:2) was used to extract carotenoids with four independent factors: ultrasound time (20–60 min); ratio of sample to solvent (1:12–1:4); saponification time (10–50 min); and concentration of saponification solution (KOH–methanol) (10–30%). A second-order polynomial model produced a satisfactory fitting of the experimental data with regard to zeaxanthin (R2 = 75.95%, p < 0.0197), lutein epoxide (R2 = 90.24%, p < 0.0001), and violaxanthin (R2 = 73.84%, p < 0.0809) content. The optimum joint extraction conditions of zeaxanthin, lutein epoxide, and violaxanthin were 40 min, 1:8, 32 min, and 20%, respectively. The optimal predicted contents for zeaxanthin (0.823022 µg/g DW), lutein epoxide (4.03684 µg/g dry; DW—dry weight), and violaxanthin (16.1972 µg/g DW) in extraction had little difference with the actual experimental values obtained under the optimum extraction conditions for each response: zeaxanthin (0.8118 µg/g DW), lutein epoxide (3.9497 µg/g DW), and violaxanthin (16.1590 µg/g DW), which provides a theoretical basis and method for cultivating new varieties at low temperatures and weak light resistance.