TAILIEUCHUNG - Recent Advances in Signal Processing 2011 Part 7

Tham khảo tài liệu 'recent advances in signal processing 2011 part 7', kỹ thuật - công nghệ, cơ khí - chế tạo máy phục vụ nhu cầu học tập, nghiên cứu và làm việc hiệu quả | Gaze prediction improvement by adding a face feature to a saliency model 197 level stimuli 14155 frames without audio signal because the model did not consider colour and audio information. Stimuli were seen in random order. Human eye position density maps The eye tracker records eye positions at 500 Hz. We recorded twenty eye positions 10 positions for each eye per frame and per subject. The median of these positions X-axis median and Y-axis median was taken for each frame and for each subject. Then for each frame we had fifteen positions one per subject . Because the final aim was to compare these positions to a saliency map a two-dimensional Gaussian was added to each position. The standard deviation at mid-height of the Gaussian was equal to of visual angle which is close to the size of the maximum resolution of the fovea. Therefore for each frame k we got a human eye position density map Mh x y k . Metric used for model evaluation We used the Normalized Scanpath Saliency NSS Peters Itti 2008 . This criterion was especially designed to compare eye fixations and the salient locations emphasized by a model saliency map. We computed the NSS metric as follows 1 NSS k Mh x y k x Mm x y k - Mm x y k 1 ơMm x y k where Mh x y k is the human eye position density map normalized to unit mean and Mm x y k a model saliency map for a frame k. The NSS is null if there is no link between eye position and salient regions. The NSS is negative if eye position tends to be in non-salient regions. The NSS is positive if eye position tends to be in salient regions. To summarize a saliency map is a good predictor of human eye fixations if the corresponding NSS value is positive and high. In the next sections we computed the NSS average over several frames. 3. The static and the dynamic pathways of the saliency model We based ourselves on the biology of the human visual system to propose a saliency model that decomposes the visual signal into a static and a dynamic .

TÀI LIỆU LIÊN QUAN
69    272    11
TỪ KHÓA LIÊN QUAN
Đã phát hiện trình chặn quảng cáo AdBlock
Trang web này phụ thuộc vào doanh thu từ số lần hiển thị quảng cáo để tồn tại. Vui lòng tắt trình chặn quảng cáo của bạn hoặc tạm dừng tính năng chặn quảng cáo cho trang web này.