Adaptive pixel intensity loss

[liuzhihui2046]

hi,thank you for your excellent work，I don't understand the following formula. Can you explain it for me？


1、Why does equation 8 need to be multiplied by yij? This will cause all background weights to be 0。
2、What is the meaning of adding 1.5 to the denominator in formula 9?
Thank you very much and hope to get your reply！！

[Karel911]

Hi @liuzhihui2046,

A1. Exactly. Before multiplying the y_ij, the omega contains not only the foreground but also background areas.
Thus, we exclude the bg area using y_ij (GTs). As a result, we can obtain the omega, as depicted in Fig 3.

A2. That is a normalization constant which coordinate a smoothness of the prediction result.
When it is small, we can obtain globally clear and smooth images (also, getting satisfactory performance on MAE metric).
In contrast, if is large, TRACER can more detect fine edges but the bg noises occur.