Algorithm 1:

Confidence level calculation for saliency maps

For saliency map i = 1:N

Find all fully connected contours with threshold > 0.3

For contour c = 1:C

$\begin{array}{l}\text{if}{I}_{argmax\left(I_{1:c}\right)}<a_1{I}_{argmax\left(I_{1:c}\right)-1}\text{and}{M}_{argmax\left(I_{1:c}\right)}<M_{argmax\left(M_{1:c}\right)}\\ \text{orif}{I}_{argmax\left(I_{1:c}\right)}<a_2{I}_{argmax\left(I_{1:c}\right)-1}\text{and}{M}_{argmax\left(I_{1:c}\right)}+a_3<M_{argmax\left(M_{1:c}\right)}\\ \text{orif}{M}_{argmax\left(M_{1:c}\right)}>a_4\text{and}argmax\left(M_{1:c}\right)\ne argmax\left(I_{1:c}\right)\end{array}$

Remove saliency map i from the set