Abstract:This paper presents a novel wavelet transform saliency model to detect salient objects. In this model, a saliency map is generated by combining orientation feature maps obtained from wavelet transform of different scale images derived from the same image. Then, the order map of a saliency map is obtained by using Fourier descriptor, which could be used as a guidance to process the most important objects. Experiments indicate that this saliency model is robust to noise and superior to other saliency models in the literature.

A Saliency Model Based on Wavelet Transform and Visual Attention