FlowEstimationbyQWT
所属分类:波变换
开发工具:matlab
文件大小:576KB
下载次数:93
上传日期:2014-07-03 21:56:07
上 传 者:
wanwandejimo
说明: 使用四元数小波变换图像流估计。是对“Coherent Multiscale Image Processing using Dual-Tree Quaternion Wavelets”的实现。代码很详细。
(Use quaternion wavelet transform image flow estimation. Is the " Coherent Multiscale Image Processing using Dual-Tree Quaternion Wavelets" implementation. Code is very detailed.)
文件列表:
FlowEstimationbyQWT (0, 2014-07-03)
FlowEstimationbyQWT\Coherent Multiscale Image Processing.pdf (601292, 2014-07-03)
FlowEstimationbyQWT\calculate_error_flows.m (1070, 2008-02-03)
FlowEstimationbyQWT\collect_phase_data.m (1676, 2005-05-18)
FlowEstimationbyQWT\combine_result_3.m (2301, 2008-02-03)
FlowEstimationbyQWT\estimate_disparity_4.m (7649, 2008-02-03)
FlowEstimationbyQWT\estimate_local_freq_4.m (1071, 2008-02-03)
FlowEstimationbyQWT\eval_flow.m (1153, 2007-02-14)
FlowEstimationbyQWT\interp_dmap.m (1475, 2008-02-03)
FlowEstimationbyQWT\pick_images.m (4634, 2008-02-03)
FlowEstimationbyQWT\read_correct_flows.m (630, 2007-02-05)
FlowEstimationbyQWT\read_sequence.m (1102, 2006-01-23)
FlowEstimationbyQWT\register_images.m (4294, 2008-02-03)
FlowEstimationbyQWT\resize_correct_flows.m (453, 2006-01-16)
FlowEstimationbyQWT\standard_QWT (0, 2014-07-03)
FlowEstimationbyQWT\standard_QWT\FSfarras.m (1740, 2004-06-22)
FlowEstimationbyQWT\standard_QWT\Icwt2.m (641, 2005-04-07)
FlowEstimationbyQWT\standard_QWT\Idwt.m (802, 1999-03-09)
FlowEstimationbyQWT\standard_QWT\Idwt2.m (2782, 2005-04-07)
FlowEstimationbyQWT\standard_QWT\Iicwt2.m (653, 2005-04-07)
FlowEstimationbyQWT\standard_QWT\Iidwt.m (936, 1999-03-09)
FlowEstimationbyQWT\standard_QWT\Iidwt2.m (2317, 2005-04-07)
FlowEstimationbyQWT\standard_QWT\Iiqwt2.m (947, 2005-04-07)
FlowEstimationbyQWT\standard_QWT\Iqwt2.m (767, 2005-04-14)
FlowEstimationbyQWT\standard_QWT\Iqwt2.m~ (767, 2005-04-14)
FlowEstimationbyQWT\standard_QWT\afb2D.m (910, 2003-09-23)
FlowEstimationbyQWT\standard_QWT\afb2D_A.m (1006, 2003-09-23)
FlowEstimationbyQWT\standard_QWT\afb3D.m (2683, 2003-10-29)
FlowEstimationbyQWT\standard_QWT\cplx_to_qtrn3D.m (1987, 2003-11-07)
FlowEstimationbyQWT\standard_QWT\cplxdual2D.m (1294, 2005-02-17)
FlowEstimationbyQWT\standard_QWT\cplxdual3D.m (1525, 2003-11-02)
FlowEstimationbyQWT\standard_QWT\cshift.m (905, 2005-05-18)
FlowEstimationbyQWT\standard_QWT\cshift2D.m (385, 2003-09-23)
FlowEstimationbyQWT\standard_QWT\cshift3D.m (634, 2003-10-29)
FlowEstimationbyQWT\standard_QWT\diagonal_qwt_plots.m (995, 2006-05-05)
FlowEstimationbyQWT\standard_QWT\draw_edge.m (701, 2005-03-26)
FlowEstimationbyQWT\standard_QWT\draw_edge.m~ (701, 2005-03-24)
FlowEstimationbyQWT\standard_QWT\draw_edge_rt.m (821, 2003-09-23)
... ...
Disparity Estimation software using Quaternion Wavelet Transform (QWT)
Written by: William Chan (Rice University)
date created: November 13, 2005
modified on Feb 3, 2008
---------------------------------------------------------
Steps to run program:
1. Please make sure you have added the correct path in Matlab so that your code can
access the "standard_QWT" directory (e.g., see line 4 in "test_register_images.m").
2. Run "test_register_images.m" to see estimation result for Rubik's cube
3. Change the first few lines in "test_register_images.m" for results of another
image sequence. May need to adjust initialization parameters for different image sequences.
---------------------------------------------------------
Matlab Files:
pick_images.m - read and resize images for qwt estimation algorithm
read_sequence.m - read image sequences from /jbarron folder
test_register_images.m - test file for register_images.m
register_images.m (main disparity estimation function)
- the following files are used in register_images.m
{
collect_phase_data.m - collect phases from QWT coefficients
estimate_local_freq_4.m - step 1 in registration (estimate effective spectral center)
estimate_disparity_4.m - step 2 (estimate disparities, unwrap phase,
combine subband/scale)
wrap_quaternion_phase.m - unwrap the first two quaternion phase angles together
interp_dmap.m - bilinearly interpolate estimation map
combine_result_3.m - final combination of all estimate results to create
one disparity map
}
read_correct_flows.m - read correct flow maps from /jbarron folder
resize_correct_flows.m - resize correct flow maps to the size of estimated map
(for easier comparison)
calculate_error_flows.m - calculate error flows given correct and estimated flows
eval_flow.m - calculate and display angular errors
----------------------------------------------------------
Other folders:
standard_QWT - code for QWT
jbarron - image sequences for testing purposes
----------------------------------------------------------
Details about the disparity estimation algorithm:
(see my thesis/journal paper)
- for usage of register_images.m, see test_register_images.m
Figure 1: Disparity estimate for a chosen scale (with QWT magnitudes,
reliability maps) for all three subbands
Remark: Can use this diagram to observe scale estimates and adjust
magnitude threshold and reliability threshold)
[see "register_images.m"] (need to first uncomment part of the code)
Figure 2: Estimated disparity flows [see "combine_result_3.m"]
Figure 3: Estimated disparity flows (overlaid on reference image)
[see "combine_result_3.m"]
Figure 4: Reference and target images (A and B) [see "test_register_images.m"]
Figure 5: Correct disparity flows (if available) [see "test_register_images.m"]
Figure 6: Error of disparity estimates (overlaid on reference image)
(if correct flows available) [see "test_register_images.m"]
- several parameters to adjust for better performance:
- magnitude threshold (in estimate_disparity_3.m)
- reliability threshold (in estimate_disparity_3.m)
- outliars threshold (in combine_result_3.m)
Remark: Unreliable estimates can attribute to small QWT coefficients,
zero-divide-by-zero in phase calculation, and occasional incorrect phase unwraps
- In this algorithm, we may observe unsatisfactory estimations near:
- boundary areas (program artificially add boundaries to make image dyadic,
see test_register_images.m)
- smooth areas (though interpolated, unreliable estimates can still cause errors)
- if underlying disparity not smooth (e.g. near image boundaries, or
cloud region in yosemite sequence)
- Remedy by getting rid of out-liars (estimates with too big magnitudes; see combine_result_3.m)
近期下载者:
相关文件:
收藏者: