Limits...
Visual attention model based on statistical properties of neuron responses.

Duan H, Wang X - Sci Rep (2015)

Bottom Line: Valuable regions are highlighted while redundant backgrounds are suppressed in the saliency maps obtained by the proposed model.Comparative results reveal that the proposed model outperforms several state-of-the-art models.This study provides insights into the neuron responses based saliency detection and may underlie the neural mechanism of early visual cortices for bottom-up visual attention.

View Article: PubMed Central - PubMed

Affiliation: 1] State Key Laboratory of Virtual Reality Technology and Systems, Beihang University, Beijing 100191, P. R. China [2] Science and Technology on Aircraft Control Laboratory, School of Automation Science and Electronic Engineering, Beihang University, Beijing 100191, P. R. China.

ABSTRACT
Visual attention is a mechanism of the visual system that can select relevant objects from a specific scene. Interactions among neurons in multiple cortical areas are considered to be involved in attentional allocation. However, the characteristics of the encoded features and neuron responses in those attention related cortices are indefinite. Therefore, further investigations carried out in this study aim at demonstrating that unusual regions arousing more attention generally cause particular neuron responses. We suppose that visual saliency is obtained on the basis of neuron responses to contexts in natural scenes. A bottom-up visual attention model is proposed based on the self-information of neuron responses to test and verify the hypothesis. Four different color spaces are adopted and a novel entropy-based combination scheme is designed to make full use of color information. Valuable regions are highlighted while redundant backgrounds are suppressed in the saliency maps obtained by the proposed model. Comparative results reveal that the proposed model outperforms several state-of-the-art models. This study provides insights into the neuron responses based saliency detection and may underlie the neural mechanism of early visual cortices for bottom-up visual attention.

Show MeSH
Well-trained connecting weight matrixes.We selected 126 connecting weight matrixes (i.e., RFs) randomly. Each square describes an oriented and localized Gabor-like RF of a specified V1 neuron. The gray pixels in each square represent zero, the lighter pixels correspond to positive values, and the darker ones indicate negative values. The localized, oriented, and band-pass RFs of neurons are somewhat like Gabor filters which is consist with that of accurate predictions of V1 RFs.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4352866&req=5

f6: Well-trained connecting weight matrixes.We selected 126 connecting weight matrixes (i.e., RFs) randomly. Each square describes an oriented and localized Gabor-like RF of a specified V1 neuron. The gray pixels in each square represent zero, the lighter pixels correspond to positive values, and the darker ones indicate negative values. The localized, oriented, and band-pass RFs of neurons are somewhat like Gabor filters which is consist with that of accurate predictions of V1 RFs.

Mentions: The connecting matrix of V1 neurons is trained before the computation of neuron responses. The E-I Net model48 is adopted in this process. 1000 patches randomly sampled from gray-scale natural images are whitened and normalized to learn Gabor-like RFs. The sparseness of neuron responses together with the error between the original and recovered images are the main factors to be considered as training rules. The size of patches is set as 14*14 (i.e., the dimension of input stimuli is 196) and 128 weight matrixes are learned finally. More details about the processing of the E-I Net model can be found in Ref. 48. An example of the learned RFs is presented in Figure 6. Neuron responses can be calculated using the learned RFs. The error between the original image and the image recovered with the corresponding neuron responses is minimized (see Result).


Visual attention model based on statistical properties of neuron responses.

Duan H, Wang X - Sci Rep (2015)

Well-trained connecting weight matrixes.We selected 126 connecting weight matrixes (i.e., RFs) randomly. Each square describes an oriented and localized Gabor-like RF of a specified V1 neuron. The gray pixels in each square represent zero, the lighter pixels correspond to positive values, and the darker ones indicate negative values. The localized, oriented, and band-pass RFs of neurons are somewhat like Gabor filters which is consist with that of accurate predictions of V1 RFs.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4352866&req=5

f6: Well-trained connecting weight matrixes.We selected 126 connecting weight matrixes (i.e., RFs) randomly. Each square describes an oriented and localized Gabor-like RF of a specified V1 neuron. The gray pixels in each square represent zero, the lighter pixels correspond to positive values, and the darker ones indicate negative values. The localized, oriented, and band-pass RFs of neurons are somewhat like Gabor filters which is consist with that of accurate predictions of V1 RFs.
Mentions: The connecting matrix of V1 neurons is trained before the computation of neuron responses. The E-I Net model48 is adopted in this process. 1000 patches randomly sampled from gray-scale natural images are whitened and normalized to learn Gabor-like RFs. The sparseness of neuron responses together with the error between the original and recovered images are the main factors to be considered as training rules. The size of patches is set as 14*14 (i.e., the dimension of input stimuli is 196) and 128 weight matrixes are learned finally. More details about the processing of the E-I Net model can be found in Ref. 48. An example of the learned RFs is presented in Figure 6. Neuron responses can be calculated using the learned RFs. The error between the original image and the image recovered with the corresponding neuron responses is minimized (see Result).

Bottom Line: Valuable regions are highlighted while redundant backgrounds are suppressed in the saliency maps obtained by the proposed model.Comparative results reveal that the proposed model outperforms several state-of-the-art models.This study provides insights into the neuron responses based saliency detection and may underlie the neural mechanism of early visual cortices for bottom-up visual attention.

View Article: PubMed Central - PubMed

Affiliation: 1] State Key Laboratory of Virtual Reality Technology and Systems, Beihang University, Beijing 100191, P. R. China [2] Science and Technology on Aircraft Control Laboratory, School of Automation Science and Electronic Engineering, Beihang University, Beijing 100191, P. R. China.

ABSTRACT
Visual attention is a mechanism of the visual system that can select relevant objects from a specific scene. Interactions among neurons in multiple cortical areas are considered to be involved in attentional allocation. However, the characteristics of the encoded features and neuron responses in those attention related cortices are indefinite. Therefore, further investigations carried out in this study aim at demonstrating that unusual regions arousing more attention generally cause particular neuron responses. We suppose that visual saliency is obtained on the basis of neuron responses to contexts in natural scenes. A bottom-up visual attention model is proposed based on the self-information of neuron responses to test and verify the hypothesis. Four different color spaces are adopted and a novel entropy-based combination scheme is designed to make full use of color information. Valuable regions are highlighted while redundant backgrounds are suppressed in the saliency maps obtained by the proposed model. Comparative results reveal that the proposed model outperforms several state-of-the-art models. This study provides insights into the neuron responses based saliency detection and may underlie the neural mechanism of early visual cortices for bottom-up visual attention.

Show MeSH