Fig. 10From: Panicle-SEG: a robust image segmentation method for rice panicles in the field based on deep learning and superpixel optimizationThe principle of entropy rate superpixel optimization. A Entropy rate superpixel image. The typical characteristic of the entropy rate superpixel image is having larger background regions and relatively small foreground panicle regions, and the algorithm has stronger abilities in edge preservation. B SLIC superpixel image. The edge of the SLIC superpixel region is relatively regulated. Sometimes, the foreground pixels and background pixels can exist in the same SLIC superpixel region, such as in the white dotted boxes (1) and (2) dpiBack to article page