Presentation is loading. Please wait.

Presentation is loading. Please wait.

100+ Times Faster Weighted Median Filter [cvpr ‘14]

Similar presentations


Presentation on theme: "100+ Times Faster Weighted Median Filter [cvpr ‘14]"— Presentation transcript:

1 100+ Times Faster Weighted Median Filter [cvpr ‘14]
Presenter: Chang-Ryeol Lee

2 Introduction Typical problems in stereo matching
Wrong disparity estimates in depth discontinuity regions Disparity image estimated by our active stereo system [samsung project ’14]

3 Introduction One of the ways to get better disparity image
Post-processing of estimated disparity image Initial disparity image Improved disparity map

4 Introduction Post processing by using color image is popular and powerful Weighted Median Filtering (WMF) Weighted Bilateral Filtering (WBF) Weighted Mode Filtering (WDF) Initial disparity image Color image Improved disparity map

5 Introduction Post processing by using color image is popular and powerful Weighted Median Filtering (WMF) Effective to salt and pepper noises Edge-preserving filter Initial disparity image Color image Improved disparity map

6 What is WMF? Based on sliding window strategies Disparity image

7 The number of disparity
What is WMF? Based on sliding window strategies Median filtering Select median value within a window 5 1 3 4 1 1 5 3 8 7 Disparity image 3 6 2 1 9 disparity The number of disparity 5 8 3 9 3 1 3 7 4 6 Median value Disparity image

8 What is WMF? Based on sliding window strategies
Weighted median filtering Select color weighted median value within a window Weight Color similarity between center point pixel and neighborhood pixel disparity Color weight I(p) I(q) D(q) Disparity image Color image 5 Weighted 히스토그램

9 What is WMF? Based on sliding window strategies
Weighted median filtering Select color weighted median value within a window Median value computation by cumulative histogram Color weight k n histogram Cumulative histogram Color weight disparity disparity

10 Limitation of WMF Nonlinearity Improvement strategies
⇒ histogram and cumulative histogram have to be computed within every window Improvement strategies Joint histogram to use box filtering technique Median tracking to cut cumulative histogram procedure Necklace table for fast access in sparse data structure median{ window1 + window2 }  median{ window1 } + median{ window2 }

11 Fast WMF Joint histogram
2D histogram composed of disparity and color intensity Color weighted histogram can be computed by multiplication of joint histogram and color similarity disparity Color intensity 255 100 d i disparity Color weight Color weighted histogram Joint histogram

12 Fast WMF Joint histogram
2D histogram composed of disparity and color intensity Only count the number of disparities and corresponding color intensity It means that we can use box filtering technique disparity 255 i Color intensity I(q) D(q) d 100 Color intensity Disparity image Joint histogram

13 Fast WMF Median tracking Insight
Colors in a window are similar to those in neighborhood window So median value is not largely changed as a window moves Tracking current median value by shifting of median value in previous window * k is median value Previous window Current window disparity Color weight disparity Color weight

14 Fast WMF Median tracking Balance
Difference between left sum and right sum of histogram disparity Color weight Left sum Right sum

15 Fast WMF .* Median tracking Fast balance computation + -
To compute balance, all values in joint histogram and corresponding color similarities should be multiplied 255 k i + - .* Color intensity Disparity 100 Joint histogram

16 Fast WMF .* Median tracking Fast balance computation + -
To compute balance, all values in joint histogram and corresponding color similarities should be multiplied 255 k i + - .* Color intensity Disparity 100 Joint histogram

17 Fast WMF Median tracking Fast balance computation + -
Balance Counting Box (BCB) Row by Row computation of balance in joint histogram Disparity Color intensity 255 100 k i + - Joint Histogram Balance Counting Box (BCB)

18 Fast WMF .* Median tracking Fast balance computation
Balance Counting Box (BCB) Multiplication of only BCB and corresponding color similarity .* Balance Counting Box (BCB)

19 Fast WMF Experiments Resolution: 640 x 480
Environments: single core, C code Comparison About 6 times faster Color image time (s) speed (fps) WMF 0.574 1.74 Fast WMF 0.091 10.98 Initial disparity image WMF Fast WMF

20 Conclusion Practical issues show up in computer vision community
The main idea of this paper can be applied to other histogram- based application

21 Box filtering


Download ppt "100+ Times Faster Weighted Median Filter [cvpr ‘14]"

Similar presentations


Ads by Google