Figure 5

Running time values of the algorithm run on images with different sizes using five different GPGPUs and the CPU as the reference. The left side displays running time obtained using the speed setting, the right side displays running times obtained using the quality setting. Each measurement displays an average value obtained from running the algorithm on 15 images. In the case of "N/A" values the onboard device memory sometimes became a bottleneck, which resulted in frequent caching operations that seriously slowed down the segmentation performance.