Enhancing Diversity of Defocus Blur Detectors via Cross-Ensemble Network

Wenda Zhao

Abstract:

Defocus blur detection (DBD) is a fundamental yet chal-
lenging topic, since the homogeneous region is obscure and
the transition from the focused area to the unfocused region
is gradual. Recent DBD methods make progress through ex-
ploring deeper or wider networks with the expense of high
memory and computation. In this paper, we propose a nov-
el learning strategy by breaking DBD problem into multi-
ple smaller defocus blur detectors and thus estimate errors
can cancel out each other. Our focus is the diversity en-
hancement via cross-ensemble network. Specifically, we de-
sign an end-to-end network composed of two logical parts:
feature extractor network (FENet) and defocus blur detec-
tor cross-ensemble network (DBD-CENet). FENet is con-
structed to extract low-level features. Then the features are
fed into DBD-CENet containing two parallel-branches for
learning two groups of defocus blur detectors. For each in-
dividual, we design cross-negative and self-negative corre-
lations and an error function to enhance ensemble diversity
and balance individual accuracy. Finally, the multiple defo-
cus blur detectors are combined with a uniformly weighted
average to obtain the final DBD map. Experimental results
indicate the superiority of our method in terms of accura-
cy and speed when compared with several state-of-the-art
methods