The goal of image segmentation is to partition an image into regions that are internally homogeneous and heterogeneous with respect to other neighbouring regions. We build on the pyramid image segmentation work proposed by [BHR] and [SSN] by introducing a mixture of colour and texture cues in order to more accurately group regions. Statistical comparison of each colour channel separately along with edge intensity and orientation histogram comparison were the cues used for region merging. The input image size is no longer constrained as in [BHR], [SSN] and other similar regular pyramid based approaches due to a modification of the pyramid construction rule. The new algorithm is tested on a set of images from the Berkeley image segmentation benchmark set. Our algorithm is fast (produces segmentations within seconds), results in the correct segmentation of elongated and large regions, very simple compared to plethora of existing algorithms, and appears competitive in segmentation quality with the best publicly available implementations.
展开▼
