We derive the delay-time distribution (DTD) of Type Ia supernovae (SNe Ia) using a sample of 132 SNe Ia, discovered by the Sloan Digital Sky Survey II (SDSS2) among 66 000 galaxies with spectral-based star formation histories (SFHs). To recover the best-fitting DTD, the SFH of every individual galaxy is compared, using Poisson statistics, to the number of SNe that it hosted (zero or one), based on the method introduced in Maoz et al. This SN sample differs from the SDSS2 SN Ia sample analysed by Brandt et al., using a related, but different, DTD recovery method. Furthermore, we use a simulation-based SN detection-efficiency function, and we apply a number of important corrections to the galaxy SFHs and SN Ia visibility times. The DTD that we find has 4σ detections in all three of its time bins: prompt (τ < 0.42 Gyr), intermediate (0.42 < τ < 2.4 Gyr) and delayed (τ > 2.4 Gyr), indicating a continuous DTD, and it is among the most accurate and precise among recent DTD reconstructions. The best-fitting power-law form to the recovered DTD is t−1.07 ± 0.07, consistent with generic ∼t−1 predictions of SN Ia progenitor models based on the gravitational-wave-induced mergers of binary white dwarfs. The time-integrated number of SNe Ia per formed stellar mass is NSN/M = 0.00130 ± 0.00015 M⊙−1, or about 4 per cent of the stars formed with initial masses in the 3 - 8 M⊙ range. This is lower than, but largely consistent with, several recent DTD estimates based on SN rates in galaxy clusters and in local-volume galaxies, and is higher than, but consistent with NSN/M estimated by comparing volumetric SN Ia rates to cosmic SFH.