Figure 9.

Pseudo-coloured image comparison between AEQE-CNN and HEVC [16] + DBF&SAO based on the absolute reconstruction error for the center SAI at position $(p,q)=(8,8),$ and for $QP=37.$ Green marks the pixel positions where AEQE-CNN achieved better performance. Blue marks the pixel positions where the two methods had the same performance. Red marks pixels where HEVC [16] + DBF&SAO achieved better performance. The cyan rectangle marks an image area shown zoomed-in at the top-left corner and the corresponding Y channel in Figure 10. The results for two LF images in the test set: (a) Chain_link_fence_2; (b) Flowers.