Improved Squeeze Algorithm for Saliency Features of Dictionary Learning
Qian Zhang, Lin Wang, Jun Cheng, Ren Long Pan · Applied Mechanics and Materials · 2014
Saliency feature is the main interesting feature of the image; we can find the goals and interesting region in the image. We proposed an improved squeeze algorithm for dictionary learning to extract significant features of an image. For the Squeeze algorithm may come into "chain effect", we solve this question by restricting its expansion radius extent dynamically, then using image sparse dictionary learning and the improved algorithm based on scene perceptions to extract the significance. Through a lot of experiments comparing between the K-SVD, Squeeze, Squeeze in [11] and our algorithm, the results show that the suggested algorithm has high robustness, anti-interference ability, and high accuracy detection.