Two-dimensional (2D) digital image analysis is efficient for investigating the rice grain shape characters in large genetic and breeding populations. In this study, we used 2D image analysis to investigate seven characters, i.e., grain length (GL), grain width (GW), length-to-width ratio (LW), grain area (GA), grain circumference (GC), grain diameter (GD), and grain roundness (GR), in one japonica × indica genetic population consisting of 215 recombinant inbred lines. GL and GW can be recorded manually as well, and have been extensively used together with LW (i.e., GL/GW) in genetic studies on grain shape. GC and GA can be hardly measured manually, and have not been used together with GD and GR. Results indicated that the seven characters could be precisely measured by 2D image analysis, genotype by environment interaction was low, and heritability was high. Each character was controlled by a few major stable genes and multiple minor additive genes. A total of 51 QTL were detected for the seven characters across four diverse environments, 22 from GL, GW, and LW, the three traditional characters, and 29 from the other four characters. The 51 QTL were clustered in eighteen marker intervals. Comparing with previous studies and analyzing the stability of identified QTL, we found six non-reported marker intervals, one each on chromosomes 2 and 3, and two each on chromosomes 6 and 8. The newly identified loci and the large-scale phenotyping system would greatly improve our knowledge about the genetic architecture and the future rice breeding on grain shape.