Some Latest Topics about Visual Attention Research in ICIP2004
ICIP2004 just finished in singapore. This is an international conference of rank 2. I pick up some papers in this conference about visual attention which is my interest research topic in my first year research. I briefly introduce these idea as follows.
1. "Towards Unsupervised Attention Object Extraction By Integrating Visual Attention and Object Growing", Junwei Han, King N. Ngan, Mingjing Li and Hongjiang Zhang
This paper proposed an object extraction algorithm using region growing and visual attention detection - "Attention Object Growing". They use attention value and local homogenerous measure to select growing seed not only be the high attention location but also near the object center. Then region growing processing to find the optimal Attention Object (AO) is modeled as MAP-MRF optimization on the parameter of threshold value.
2. "Performance Assessment of A Visual Attention System Entirely Based on Human Vision Modeling", O. Le Meur, P. Le callet, D. Barba and D. Thoreau
This paper proposed the Subjective Comparison using eye tracking data and Objective Comparison using lullback-leibler divergence to access the performance of visual attention models. From their analysis, they point out human attention depends on the viewing time and their visual attention outperform to Itti's reference model 12% and similar to the human vision.
3. "The Saliency Grouping Field", Maurizio Pilu (HP Labs Europe)
How to find organization in real saliency map is addressed in this paper. They used orientation selective kernels to convolve with the maxima salient locations to find the initial grouping field. A reinforcing process is then applied to improlve perceptually salient component (orientation). With this grouping field, they claim some useful structure in the image can be elicited.
1. "Towards Unsupervised Attention Object Extraction By Integrating Visual Attention and Object Growing", Junwei Han, King N. Ngan, Mingjing Li and Hongjiang Zhang
This paper proposed an object extraction algorithm using region growing and visual attention detection - "Attention Object Growing". They use attention value and local homogenerous measure to select growing seed not only be the high attention location but also near the object center. Then region growing processing to find the optimal Attention Object (AO) is modeled as MAP-MRF optimization on the parameter of threshold value.
2. "Performance Assessment of A Visual Attention System Entirely Based on Human Vision Modeling", O. Le Meur, P. Le callet, D. Barba and D. Thoreau
This paper proposed the Subjective Comparison using eye tracking data and Objective Comparison using lullback-leibler divergence to access the performance of visual attention models. From their analysis, they point out human attention depends on the viewing time and their visual attention outperform to Itti's reference model 12% and similar to the human vision.
3. "The Saliency Grouping Field", Maurizio Pilu (HP Labs Europe)
How to find organization in real saliency map is addressed in this paper. They used orientation selective kernels to convolve with the maxima salient locations to find the initial grouping field. A reinforcing process is then applied to improlve perceptually salient component (orientation). With this grouping field, they claim some useful structure in the image can be elicited.
