Water chestnut insoluble dietary fiber (WIDF) was prepared by Ultrasound-assisted enzymatic extraction from the by-product of starch-type water chestnut, dregs, as the raw material. The effects of composite enzyme dosage, solid-liquid ratio and ultrasonic power on the extraction rate of WIDF were investigated by single factor experiments. Response surface experiments were designed and performed to optimize the extraction process, the main factors affecting the indices and the interactions among various factors were analyzed, and quadratic regression models were established. The obtained optimal process parameters were optimized and verified, and comparisons on the quality and functional characteristics between the WIDF samples extracted by the traditional composite enzyme method and the optimized method were made. The results showed that the established regression model of WIDF extraction rate was extremely significant (P<0.01), thus, the model could be used to analyze and predict the extraction rate for the WIDF obtained by the ultrasound-assisted enzymatic extraction method. The optimal parameters obtained after process optimizationwere as follows: enzyme addition amount 9%, solid-liquid ratio 1:15, and ultrasonic power 320 W. Verification tests were carried out under such conditions, and the extraction rate of WIDF was as high as 67.60%. Compared with the traditional composite enzyme method, the extraction rate of WIDF under optimized conditions increased by 107.36%, with WIDF’s water-holding capacity of WIDF increasing by 34.33% and cholesterol absorption performance increasing by 35.77%. This research provides a theoretical basis for the high-value comprehensive utilization of water chestnut dregs.