Method

Hierarchical coherency sensitive hashing and interpolation with RANSAC [HCSH]


Submitted on 25 Dec. 2019 03:22 by
Jingzhe Fan (Beihang University)

Running time:3.5 s
Environment:1 core @ 3.0 Ghz (C/C++)

Method Description:
Perform NNF with coarse-to-fine scheme to improve
coherency.

Filter matches with simple saliency as well as
forward–backward constant.

Locally affine estimation with RANSAC to increase the
robustness of interpolation.
Parameters:
[PatchMatchCtF]
verbose=0
limitHoriz=0
useDAISY=1
gridSize=3
borderSize=5
maxLevelIteration=9
downFactor=0.5f
sigma=0.8f
minSize=30
maxDisplacement=160.0f
stopItRatio=0.0001f
fbThreshold=3

[EdgeAwareInterpolator]
verbose=0
saliencyFilter=1
preFilter=1
useCity=1
useLA=1
useRanSaC=1
useVarRefine=1
saliencyTh=0.06f
preNNK=32
preTh=5.0f
nnK=160
euc=0.001f
kappa=-1.0f
ransacIteration=5
dtIteration=1
regularCoef=0.01f

[VariationalRefinement]
verbose=0
variationalIteration=5
fixedPointIteration=1
sorIteration=15
omega=1.6f
alpha=1.0f
beta=-1.0f
delta=0.2f
gamma=0.6f
kappa=-1.0f
zeta=0.1f
epsilon=0.001f
Latex Bibtex:
@article{FAN20181,
title = "Hierarchical coherency sensitive hashing and interpolation with
RANSAC for large displacement optical flow",
journal = "Computer Vision and Image Understanding",
volume = "175",
pages = "1 - 10",
year = "2018",
issn = "1077-3142",
doi = "https://doi.org/10.1016/j.cviu.2018.10.005",
url =
"http://www.sciencedirect.com/science/article/pii/S1077314218304223",
author = "Jingzhe Fan and Yan Wang and Lei Guo",
keywords = "Nearest neighbor field, Sparse to dense interpolation,
Optical flow",
abstract = "Nearest Neighbor Field (NNF) has shown excellent performance
for large displacement optical flow estimation recently. However, it
contains much noise and lacks of global constraint. In this paper we
present an effective approach, named HCSH (Hierarchical Coherency
Sensitive Hashing), which combines the coarse-to-fine scheme and random
search strategy, to enable NNF to enjoy the inherent smooth of coarse-
to-fine framework. Then besides the forward–backward check for NNF, we
also use auto-correlation to remove the unreliable correspondences in
flat regions, where the NNF is often considered ambiguous and the motion
can be naturally recovered by the latter interpolation. Inspired by
EpicFlow, we propose edge-aware interpolation (EAI-Flow) to filling the
gaps by removing correspondence. RANSAC is introduced to improve the
locality-weighted affine transformation estimation, with neighbor
propagation of affine model to reduce required iterations and speed up
the computation. Experimental validation shows that our approach
outperforms the state-of-the-art with more accurate optical flow."
}

Detailed Results

This page provides detailed results for the method(s) selected. For the first 20 test images, the percentage of erroneous pixels is depicted in the table. We use the error metric described in Object Scene Flow for Autonomous Vehicles (CVPR 2015), which considers a pixel to be correctly estimated if the disparity or flow end-point error is <3px or <5% (for scene flow this criterion needs to be fulfilled for both disparity maps and the flow map). Underneath, the left input image, the estimated results and the error maps are shown (for disp_0/disp_1/flow/scene_flow, respectively). The error map uses the log-color scale described in Object Scene Flow for Autonomous Vehicles (CVPR 2015), depicting correct estimates (<3px or <5% error) in blue and wrong estimates in red color tones. Dark regions in the error images denote the occluded pixels which fall outside the image boundaries. The false color maps of the results are scaled to the largest ground truth disparity values / flow magnitudes.

Test Set Average

Error Fl-bg Fl-fg Fl-all
All / All 18.05 26.23 19.41
All / Est 18.05 26.23 19.41
Noc / All 9.39 22.05 11.69
Noc / Est 9.39 22.05 11.69
This table as LaTeX

Test Image 0

Error Fl-bg Fl-fg Fl-all
All / All 5.33 46.05 10.91
All / Est 5.33 46.05 10.91
Noc / All 4.63 46.05 10.91
Noc / Est 4.63 46.05 10.91
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 1

Error Fl-bg Fl-fg Fl-all
All / All 6.22 43.06 10.34
All / Est 6.22 43.06 10.34
Noc / All 4.09 43.06 8.90
Noc / Est 4.09 43.06 8.90
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 2

Error Fl-bg Fl-fg Fl-all
All / All 12.97 44.47 14.50
All / Est 12.97 44.47 14.50
Noc / All 6.75 44.47 8.94
Noc / Est 6.75 44.47 8.94
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 3

Error Fl-bg Fl-fg Fl-all
All / All 21.31 96.48 28.26
All / Est 21.31 96.48 28.26
Noc / All 10.72 95.30 18.00
Noc / Est 10.72 95.30 18.00
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 4

Error Fl-bg Fl-fg Fl-all
All / All 28.14 35.80 29.41
All / Est 28.14 35.80 29.41
Noc / All 17.69 35.02 20.93
Noc / Est 17.69 35.02 20.93
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 5

Error Fl-bg Fl-fg Fl-all
All / All 19.28 0.00 17.55
All / Est 19.28 0.00 17.55
Noc / All 13.01 0.00 11.67
Noc / Est 13.01 0.00 11.67
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 6

Error Fl-bg Fl-fg Fl-all
All / All 15.36 2.52 14.01
All / Est 15.36 2.52 14.01
Noc / All 12.64 2.52 11.44
Noc / Est 12.64 2.52 11.44
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 7

Error Fl-bg Fl-fg Fl-all
All / All 1.04 90.26 18.50
All / Est 1.04 90.26 18.50
Noc / All 1.04 89.70 17.61
Noc / Est 1.04 89.70 17.61
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 8

Error Fl-bg Fl-fg Fl-all
All / All 0.72 18.48 4.00
All / Est 0.72 18.48 4.00
Noc / All 0.72 18.48 4.00
Noc / Est 0.72 18.48 4.00
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 9

Error Fl-bg Fl-fg Fl-all
All / All 0.95 40.13 10.96
All / Est 0.95 40.13 10.96
Noc / All 0.95 40.13 10.96
Noc / Est 0.95 40.13 10.96
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 10

Error Fl-bg Fl-fg Fl-all
All / All 8.13 16.66 10.08
All / Est 8.13 16.66 10.08
Noc / All 5.85 16.66 8.66
Noc / Est 5.85 16.66 8.66
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 11

Error Fl-bg Fl-fg Fl-all
All / All 9.88 13.94 10.61
All / Est 9.88 13.94 10.61
Noc / All 5.59 13.84 7.25
Noc / Est 5.59 13.84 7.25
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 12

Error Fl-bg Fl-fg Fl-all
All / All 12.06 7.11 11.72
All / Est 12.06 7.11 11.72
Noc / All 3.63 7.11 3.91
Noc / Est 3.63 7.11 3.91
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 13

Error Fl-bg Fl-fg Fl-all
All / All 11.75 31.01 14.12
All / Est 11.75 31.01 14.12
Noc / All 3.08 21.48 5.22
Noc / Est 3.08 21.48 5.22
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 14

Error Fl-bg Fl-fg Fl-all
All / All 12.30 14.62 12.34
All / Est 12.30 14.62 12.34
Noc / All 3.75 14.62 3.98
Noc / Est 3.75 14.62 3.98
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 15

Error Fl-bg Fl-fg Fl-all
All / All 27.65 17.29 26.71
All / Est 27.65 17.29 26.71
Noc / All 11.07 17.29 11.79
Noc / Est 11.07 17.29 11.79
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 16

Error Fl-bg Fl-fg Fl-all
All / All 30.55 22.26 29.33
All / Est 30.55 22.26 29.33
Noc / All 14.80 22.26 16.14
Noc / Est 14.80 22.26 16.14
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 17

Error Fl-bg Fl-fg Fl-all
All / All 23.01 33.55 24.11
All / Est 23.01 33.55 24.11
Noc / All 8.85 33.55 12.05
Noc / Est 8.85 33.55 12.05
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 18

Error Fl-bg Fl-fg Fl-all
All / All 32.64 100.00 64.64
All / Est 32.64 100.00 64.64
Noc / All 13.96 100.00 37.69
Noc / Est 13.96 100.00 37.69
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 19

Error Fl-bg Fl-fg Fl-all
All / All 29.08 16.23 27.62
All / Est 29.08 16.23 27.62
Noc / All 13.97 16.23 14.30
Noc / Est 13.97 16.23 14.30
This table as LaTeX

Input Image

Flow Result

Flow Error




eXTReMe Tracker