Automation technologies for bin-picking are increasingly sought after in industrial applications related to material handling and order fulfillment. In this paper, we present a model-free approach to robotic bin-picking which is designed to be highly adaptable to varying imaging conditions such as lighting, rotations, scaling, shadows, etc., and highly flexible to large number of items. The proposed object imaging approach is based on visual registration through color over-segmentation with depth imaging correction. Furthermore, we propose a computationally efficient object grasp pose estimation algorithm based on planar constrained geometry on the manipulator and imaging system.We validated the performance of our imaging method using publicly available RGBD data sets, and we measured the timing and grasp repeatability using physical experiments conducted with an industrial manipulator. Results show that our grasp success rates are comparable to recently published methods, but our grasp computation speeds are considerably faster. In particular, measured image processing and grasp calculation times are of the order of 300 fps for 320x240 image size, and will scale linearly with the imaging area.