An unsupervised rough cognition algorithm for salient object extraction

An unsupervised rough cognition algorithm based on granular computing for salient object extraction is proposed in this paper. Firstly, a granular computing model, as a way to simulate human thinking, is defined. Then, the following steps are done successively: 1) Partition the universe according to two concepts respectively, and filter topology equivalence classes with smaller scale; 2) Quantize the significance of topology equivalence classes with topology connectivity and topology distribution density; 3) Find the cut-off position in the significance sequence with the improved Fisher's linear discriminant algorithm, and then obtain the candidate regions by removing non-significant topology equivalence classes; 4) Express the gradual changing pattern with dimensional scan gradient, which is used to do the local rough segmentation until candidate objects are available; merge the candidate objects that follow the gradual changing pattern and refresh the significant values of these candidate objects; 5) Run the Fisher's linear discriminant algorithm again, and determine the final salient object according to position weight if more than one candidate object is left. Finally, the executing process of the proposed approach is validated by experiment step by step, and a comparative analysis with three recent methods is conducted, which shows the superiority of the proposed approach in terms of ability to approximate semantic of object and speed.