Cpfind - Revision history

Thomas: Removed now obsolete warning

2012-10-15T16:22:16Z

Removed now obsolete warning

Tve: minor clarification

2012-10-15T00:07:54Z

minor clarification

Thomas: /* Matching strategy */

2012-03-18T09:16:42Z

‎Matching strategy

Thomas: Typos

2011-09-18T12:07:45Z

Typos

Thomas: Clarify some details and fixes some typos

2011-08-31T17:03:47Z

Clarify some details and fixes some typos

Jwatters: ransacmodel to ransacmode

2011-04-15T20:14:12Z

ransacmodel to ransacmode

Beeswax: /* Feature matching */ linked to wikipedia RANSAC article

2011-03-26T15:45:41Z

‎Feature matching: linked to wikipedia RANSAC article

Thomas: Added ransac mode

2011-01-30T18:18:28Z

Added ransac mode

Pablo: /* Feature matching */

2011-01-21T08:59:18Z

‎Feature matching

Pablo: /* Feature description */ clarified sieve1 parameters.

2011-01-21T08:58:28Z

‎Feature description: clarified sieve1 parameters.

@@ Line 9: / Line 9: @@
 == Rectilinear and fisheye images ==
 Cpfind can find control points in rectilinear and fisheye images. To achieve good control points images with a high horizontal field of view (e.g. ultra wide rectilinear or fisheye) are therefore remapped into a conformal space (cpfind is using the [[Projection#Stereographic_projection|stereographic projection]]) and the feature matching occurs in this space. Before writing the control points the coordinates are remapped back to the image space. This happens automatic depending on the information about the lens in the input project file. So check that your input project file contains reasonable information about the used lens.
 == Using celeste ==

@@ Line 37: / Line 37: @@
 This is an optimized matching strategy for single and multi-row panorama:
      cpfind --multirow -o output.pto input.pto
-The algorithm is the same as described in [[Hugin_Parameters_for_Control_Point_Detectors_dialog#Multi-row_panorama|multi-row panorama]]. By integrating this algorithm into '''cpfind''' it is faster by using several cores of modern CPUs and don't caching the keypoints to disc (which is time consuming). If you want to use this multi-row matching inside [[hugin]] set the control point detector type to [[Hugin_Parameters_for_Control_Point_Detectors_dialog#All_images_at_once|All images at once]].
+The algorithm is the same as described in [[Hugin_Parameters_for_Control_Point_Detectors_dialog#Multi-row_panorama|multi-row panorama]]. By integrating this algorithm into '''cpfind''' it is faster by using several cores of modern CPUs and don't caching the keypoints to disc (which is time consuming). If you want to use this multi-row matching inside [[hugin]] set the control point detector type to [[Hugin_Parameters_for_Control_Point_Detectors_dialog#All_images_at_once|All images at once]], this is so that Hugin passes all images to cpfind and lets cpfind apply the multirow matching smarts itself.
 === Matching overlapping images (prealigned panorama) ===

@@ Line 38: / Line 38: @@
      cpfind --multirow -o output.pto input.pto
 The algorithm is the same as described in [[Hugin_Parameters_for_Control_Point_Detectors_dialog#Multi-row_panorama|multi-row panorama]]. By integrating this algorithm into '''cpfind''' it is faster by using several cores of modern CPUs and don't caching the keypoints to disc (which is time consuming). If you want to use this multi-row matching inside [[hugin]] set the control point detector type to [[Hugin_Parameters_for_Control_Point_Detectors_dialog#All_images_at_once|All images at once]].
 === Keypoints caching to disc ===

@@ Line 85: / Line 85: @@
 --ransacmode <string>        Select the mode used in the ransac step. Possible values: auto, hom, rpy, rpyv, rpyb (default: auto)
-* hom: Assume a homography. Only applicable for non-wide angle views. Uses the original panomatic code. It is also more flexible than required and can generate false matches, particlarly if most of the matches are located on a single line.
+* hom: Assume a homography. Only applicable for non-wide angle views. Uses the original panomatic code. It is also more flexible than required and can generate false matches, particularly if most of the matches are located on a single line.
-* rpy: Align images using roll, pitch and yaw. This requires a good estimate for the horizontal field of view (and distortion, for  heavily distorted images). It is the prefered mode if a calibrated lens is used, or the HFOV could be read successfully from the EXIF data.
+* rpy: Align images using roll, pitch and yaw. This requires a good estimate for the horizontal field of view (and distortion, for  heavily distorted images). It is the preferred mode if a calibrated lens is used, or the HFOV could be read successfully from the EXIF data.
 * rpyv: Align pair by optimizing roll, pitch, yaw and field of view. Should work without prior knowledge of the field of view, but might fail more often, due to error function used in the panotools optimizer, it tends to shrink the fov to 0.
-* rpyvb: Align pair by optimizing roll, pitch, yaw, field of view and the "b" distortion parameter.  Probaby very fragile, just implemented for testing.
+* rpyvb: Align pair by optimizing roll, pitch, yaw, field of view and the "b" distortion parameter.  Probably very fragile, just implemented for testing.
 * auto: Use homography for images with hfov < 65 degrees and rpy otherwise.

@@ Line 3: / Line 3: @@
        cpfind -o output.pto input.pto
 Internal the control point detector algorithm is divided into 2 parts:
-* The first step is the feature description: In this step the images of the project file are loaded and so called keypoints are searched. They describe destinctive features in the image. Cpfind is using a gradient based descriptor for the feature description of the keypoints.
+* The first step is the feature description: In this step the images of the project file are loaded and so called keypoints are searched. They describe distinctive features in the image. Cpfind is using a gradient based descriptor for the feature description of the keypoints.
-* In a second step, the feature matching, all keypoints of two images are matched against each other to find features which are on both images. If this matching was successfull two keypoints in the two images become one control point.
+* In a second step, the feature matching, all keypoints of two images are matched against each other to find features which are on both images. If this matching was successful two keypoints in the two images become one control point.
 = Usage =
 == Rectilinear and fisheye images ==
-Cpfind can find control points in rectilinear and fisheye images. To achieve good control points images with a high horizontal field of view (e.g. ultra wide rectilinear or fisheye) are therefor remapped into a conformal space (cpfind is using the [[Projection#Stereographic_projection|stereographic projection]]) and the feature matching occurs in this space. Before writing the control points the coordinates are remapped back to the image space. This happens automatic depending on the information about the lens in the input project file. So check that your input project file contains reasonable information about the used lens.
+Cpfind can find control points in rectilinear and fisheye images. To achieve good control points images with a high horizontal field of view (e.g. ultra wide rectilinear or fisheye) are therefore remapped into a conformal space (cpfind is using the [[Projection#Stereographic_projection|stereographic projection]]) and the feature matching occurs in this space. Before writing the control points the coordinates are remapped back to the image space. This happens automatic depending on the information about the lens in the input project file. So check that your input project file contains reasonable information about the used lens.
 '''WARNING:''' cpfind in hugin 2010.4 has some bugs and won't work properly with fisheye images. Use a 2010.5 development snapshot (rev c137328f1418, from 2011-01-20) or wait for the next release.
@@ Line 45: / Line 45: @@
      cpfind -k 0 -k 1 input.pto
 The keypoint files are saved by default into the same directory as the images with the extension .key. In this case no matching of images occurs and therefore no output project file needs to specified. If '''cpfind''' finds keyfiles for an image in the project it will use them automatically and not run the feature descriptor again on this image.
-If you want to save them to annother directory use the --keypath switch.
+If you want to save them to another directory use the --keypath switch.
 This procedure can also be automate with the switch --cache:
@@ Line 73: / Line 73: @@
 '''Cpfind''' stores maximal sieve1width * sieve1height * sieve1size keypoints per image. If you have only a small overlap, e.g. for 360 degree panorama shoot with fisheye images, you can get better results if you increase sieve1size. You can also try to increase sieve1width and/or sieve1height.
-Effectively cpfind splits your image in rectangles: sieve1width horizontally by sieve1height vertically. It will try to find sieve1size interest points. This ensures a reasonably uniform distribution of interest points over your image. These features will be matched in the matching step. With the default parameters, up to 10000 interest points will be used in the feature matching step.
+Effectively cpfind splits your image in rectangles or buckets: sieve1width horizontally by sieve1height vertically. It will try to find sieve1size keypoints per bucket. This ensures a reasonably uniform distribution of interest points over your image. These features will be matched in the matching step. With the default parameters, up to 10000 interest points will be used in the feature matching step.
 == Feature matching ==
@@ Line 105: / Line 105: @@
 '''Cpfind''' generates between minmatches and sieve2width * sieve2height * sieve2size control points between an image pair. (Default setting is between 4 and 25 (=5*5*1) control points per image pair.)
 If less then minmatches control points are found for a given image pairs these control points are disregarded and this image pair is considered as not connected. For narrow overlaps you can try to decrease minmatches, but this increases the risk of getting wrong control points.
 [[Category:Software:Hugin]]

@@ Line 83: / Line 83: @@
 --ransacdist <int>            Ransac: homography estimation distance threshold (pixels) (default: 25)
---ransacmodel <string>        Select the model used in the ransac step. Possible values: auto, hom, rpy, rpyv, rpyb
+--ransacmode <string>        Select the mode used in the ransac step. Possible values: auto, hom, rpy, rpyv, rpyb (default: auto)
 * hom: Assume a homography. Only applicable for non-wide angle views. Uses the original panomatic code. It is also more flexible than required and can generate false matches, particlarly if most of the matches are located on a single line.

@@ Line 79: / Line 79: @@
 Fine-tuning of the matching step by the following parameters:
---ransaciter <int>            Ransac: iterations (default: 1000)
+--ransaciter <int>            [http://en.wikipedia.org/wiki/RANSAC Ransac]: iterations (default: 1000)
 --ransacdist <int>            Ransac: homography estimation distance threshold (pixels) (default: 25)

← Older revision		Revision as of 18:18, 30 January 2011
Line 82:		Line 82:

	--ransacdist <int> Ransac: homography estimation distance threshold (pixels) (default: 25)		--ransacdist <int> Ransac: homography estimation distance threshold (pixels) (default: 25)
		+
		+	--ransacmodel <string> Select the model used in the ransac step. Possible values: auto, hom, rpy, rpyv, rpyb
		+
		+	* hom: Assume a homography. Only applicable for non-wide angle views. Uses the original panomatic code. It is also more flexible than required and can generate false matches, particlarly if most of the matches are located on a single line.
		+
		+	* rpy: Align images using roll, pitch and yaw. This requires a good estimate for the horizontal field of view (and distortion, for heavily distorted images). It is the prefered mode if a calibrated lens is used, or the HFOV could be read successfully from the EXIF data.
		+
		+	* rpyv: Align pair by optimizing roll, pitch, yaw and field of view. Should work without prior knowledge of the field of view, but might fail more often, due to error function used in the panotools optimizer, it tends to shrink the fov to 0.
		+
		+	* rpyvb: Align pair by optimizing roll, pitch, yaw, field of view and the "b" distortion parameter. Probaby very fragile, just implemented for testing.
		+
		+	* auto: Use homography for images with hfov < 65 degrees and rpy otherwise.

	--minmatches <int> Minimum matches (default: 4)		--minmatches <int> Minimum matches (default: 4)

@@ Line 89: / Line 89: @@
 --sieve2height <int>          Sieve 2: Number of buckets on height (default: 5)
---sieve2size <int>            Sieve 2: Max points per bucket (default: 2)
+--sieve2size <int>            Sieve 2: Max points per bucket (default: 1)
-'''Cpfind''' generates between minmatches and sieve2width * sieve2height * sieve2size control points between an image pair. (Default setting is between 4 and 50 (=5*5*2) control points per image pair.)
+'''Cpfind''' generates between minmatches and sieve2width * sieve2height * sieve2size control points between an image pair. (Default setting is between 4 and 25 (=5*5*1) control points per image pair.)
 If less then minmatches control points are found for a given image pairs these control points are disregarded and this image pair is considered as not connected. For narrow overlaps you can try to decrease minmatches, but this increases the risk of getting wrong control points.
 [[Category:Software:Hugin]]

@@ Line 65: / Line 65: @@
 --sieve1height <int>          Sieve 1: Number of buckets on height (default: 10)
---sieve1size <int>            Sieve 1: Max points per bucket (default: 30)
+--sieve1size <int>            Sieve 1: Max points per bucket (default: 100)
---kdtreesteps <int>           KDTree: search steps (default: 40)
+--kdtreesteps <int>           KDTree: search steps (default: 200)
---kdtreeseconddist <double>   KDTree: distance of 2nd match (default: 0.15)
+--kdtreeseconddist <double>   KDTree: distance of 2nd match (default: 0.25)
 '''Cpfind''' stores maximal sieve1width * sieve1height * sieve1size keypoints per image. If you have only a small overlap, e.g. for 360 degree panorama shoot with fisheye images, you can get better results if you increase sieve1size. You can also try to increase sieve1width and/or sieve1height.
-Effectively cpfind splits your image in rectangles: sieve1width horizontally by sieve1height vertically. It will try to find sieve1size control points in each area. This ensures a reasonably uniform distribution of keypoints over your image. Of course with the defaults, a very large number (up to 3000) of controlpoints could be generated (if two images overlap fully).
+Effectively cpfind splits your image in rectangles: sieve1width horizontally by sieve1height vertically. It will try to find sieve1size interest points. This ensures a reasonably uniform distribution of interest points over your image. These features will be matched in the matching step. With the default parameters, up to 10000 interest points will be used in the feature matching step.
 == Feature matching ==