Survey on Various Approaches of Saliency Detection

Saliency detection is one of the most attractive and active topics in the field of computer vision. Just as the name suggests, saliency detection is an idea to find out the most impressive part of an image or a video. Although it contributes to many of nowadays technology, the idea of Saliency detection appeared as early as the end of 20th century with the introduction of the Itti Model. However, with a 20-year development, current method of deep learning becomes an important part of researching on saliency detection. With the participation of convolutional neutral network, the accuracy and efficiency of saliency detection has improved a lot recently. In this paper at first, some classic traditional methods (without deep learning) are introduced first. Then, there will be some more developed recent algorithms which adjust with the idea of convolutional neutral network. Finally, an evaluation of both kinds of methods as well as a brief conclusion will be given.