Visual saliency detection has lately witnessed substantial progress attributed to powerful feature representation leveraging deep convolutional neural networks (CNNs). However, existing CNN-based method has a lot of redundant computation resulting in inferring saliency maps is very time-consuming. In this paper, we propose a multiscale contrast regions deep learning framework employed to calculate salient score of an integrated image. Experimental results demonstrate that our approach is capable of achieving almost the same performance on the four public benchmarks compared to the relevant method MDF. Meanwhile, the computational efficiency is remarkably improved, when inferring the image of 400 * 300 size only takes average 3.32 s using our algorithm while MDF method consumes 8.0 s reducing rough 60% cost.
展开▼