-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
move subclasses of stitcher back to stitcher module.
- Loading branch information
Showing
4 changed files
with
201 additions
and
208 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,200 +0,0 @@ | ||
| import cv2 | ||
| import numpy as np | ||
|
|
||
| import bb_stitcher.helpers as helpers | ||
| import bb_stitcher.picking.picker as picker | ||
| from bb_stitcher.stitcher import Stitcher | ||
|
|
||
|
|
||
| class FeatureBasedStitcher(Stitcher): | ||
| """Class to create a feature based stitcher.""" | ||
|
|
||
| def __init__(self, config=None, rectify=True): | ||
| """Initialize a feature based stitcher. | ||
| Args: | ||
| config: config file which holds the basic stitching parameters. | ||
| """ | ||
| # TODO(gitmirgut) initialize super() | ||
| if config is None: | ||
| self.config = helpers.get_default_config() | ||
| else: | ||
| self.config = config | ||
| super().__init__(self.config, rectify) | ||
| self.overlap = int(self.config['FeatureBasedStitcher']['OVERLAP']) | ||
| self.border_top = int(self.config['FeatureBasedStitcher']['BORDER_TOP']) | ||
| self.border_bottom = int(self.config['FeatureBasedStitcher']['BORDER_BOTTOM']) | ||
| self.transform = self.config['FeatureBasedStitcher']['TRANSFORM'] | ||
| self.hessianThreshold = float(self.config['SURF']['HESSIANTHRESHOLD']) | ||
| self.nOctaves = int(self.config['SURF']['N_OCTAVES']) | ||
| self.max_shift_y = int(self.config['FeatureMatcher']['MAX_SCHIFT_Y']) | ||
|
|
||
| @staticmethod | ||
| def _calc_feature_mask(size_left, size_right, overlap, border_top, border_bottom): | ||
| """Calculate the masks, which defines the area for feature detection. | ||
| The mask is used to shrink the area for searching features. | ||
| Args: | ||
| size_left (tuple): Size of the left image. | ||
| size_right (tuple): Size of the right image. | ||
| overlap (int): Estimated overlap of both images in px. | ||
| border_top (int): Estimated border size on top of both images in px. | ||
| border_bottom (int): Estimated border size on top of both images in px. | ||
| Returns: | ||
| - **mask_left** (ndarray) -- mask area of the left image to search for features. | ||
| - **mask_right** (ndarray) -- mask area of the right image to search for features. | ||
| """ | ||
| # TODO(gitmirgut): Add note, why to use border. | ||
| mask_left = np.zeros(size_left[:: - 1], np.uint8) | ||
| mask_right = np.zeros(size_right[:: - 1], np.uint8) | ||
| mask_left[border_top:size_left[1] - border_bottom, size_left[0] - overlap:] = 255 | ||
| mask_right[border_top:size_left[1] - border_bottom, :overlap] = 255 | ||
|
|
||
| return mask_left, mask_right | ||
|
|
||
| def estimate_transform(self, image_left, image_right, angle_left=0, angle_right=0): | ||
| """Estimate transformation for stitching of images based on feature matching. | ||
| Args: | ||
| image_left (ndarray): Input left image. | ||
| image_right (ndarray): Input right image. | ||
| angle_left (int): Angle in degree to rotate left image. | ||
| angle_right (int): Angle in degree to rotate right image. | ||
| Returns: | ||
| - **homo_left** (ndarray) -- homography *(3,3)* for ``image_left`` to form a panorama. | ||
| - **homo_right** (ndarray) -- homography *(3,3)* for ``image_right`` to form a panorama. | ||
| - **pano_size** (tuple) -- Size *(width, height)* of the panorama. | ||
| """ | ||
| self.size_left = image_left.shape[:2][::-1] | ||
| self.size_right = image_right.shape[:2][::-1] | ||
|
|
||
| # rectify and rotate images | ||
| image_left, affine_left = self._prepare_image(image_left, angle_left) | ||
| image_right, affine_right = self._prepare_image(image_right, angle_right) | ||
|
|
||
| rot_size_left = image_left.shape[:2][:: - 1] | ||
| rot_size_right = image_right.shape[:2][:: - 1] | ||
|
|
||
| # calculates the mask which will mark the feature searching area. | ||
| mask_left, mask_right = self._calc_feature_mask( | ||
| rot_size_left, rot_size_right, self.overlap, self.border_top, self.border_bottom) | ||
|
|
||
| # Initialize the feature detector and descriptor SURF | ||
| # http://www.vision.ee.ethz.ch/~surf/download.html | ||
| # is noncommercial licensed | ||
| # TODO(gitmirgut) add note to doc and requirements on license | ||
| surf = cv2.xfeatures2d.SURF_create( | ||
| hessianThreshold=self.hessianThreshold, nOctaves=self.nOctaves) | ||
| surf.setUpright(True) | ||
| surf.setExtended(128) | ||
|
|
||
| kps_left, ds_left = surf.detectAndCompute(image_left, mask_left) | ||
| kps_right, ds_right = surf.detectAndCompute(image_right, mask_right) | ||
|
|
||
| assert (len(kps_left) > 0 and len(kps_right) > 0) | ||
|
|
||
| # Start with Feature Matching | ||
| bf = cv2.BFMatcher() | ||
|
|
||
| # search the 2 best matches for each left descriptor (ds_left) | ||
| raw_matches = bf.knnMatch(ds_left, ds_right, k=2) | ||
|
|
||
| good_matches = [] | ||
| for m in raw_matches: | ||
| if len(m) == 2 and m[0].distance < m[1].distance: | ||
| good_match = m[0] | ||
| keypoint_left = np.array(kps_left[good_match.queryIdx].pt) | ||
| keypoint_right = np.array(kps_right[good_match.trainIdx].pt) | ||
| dist = abs(keypoint_left - keypoint_right) | ||
|
|
||
| # checks if the distance in the y direction is to big | ||
| if dist[1] < self.max_shift_y: | ||
| good_matches.append(good_match) | ||
|
|
||
| good_pts_left = np.float32( | ||
| [kps_left[m.queryIdx].pt for m in good_matches]).reshape(-1, 1, 2) | ||
| good_pts_right = np.float32( | ||
| [kps_right[m.trainIdx].pt for m in good_matches]).reshape(-1, 1, 2) | ||
|
|
||
| assert len(good_matches) > 2 | ||
| homo_right = cv2.estimateRigidTransform(good_pts_left, good_pts_right, False) | ||
| if homo_right is None: | ||
| return None | ||
| homo_right = cv2.invertAffineTransform(homo_right) | ||
| homo_right = np.vstack([homo_right, [0, 0, 1]]) | ||
|
|
||
| # include the previous rotation | ||
| homo_left = affine_left | ||
| homo_right = homo_right.dot(affine_right) | ||
|
|
||
| homo_trans, pano_size = helpers.align_to_display_area( | ||
| rot_size_left, rot_size_right, homo_left, homo_right) | ||
|
|
||
| self.homo_left = homo_trans.dot(homo_left) | ||
| self.homo_right = homo_trans.dot(homo_right) | ||
| self.pano_size = pano_size | ||
|
|
||
|
|
||
| class RectangleStitcher(Stitcher): | ||
| """Class to create a rectangle stitcher. | ||
| The ``RectangleStitcher`` maps selected points to an abstracted rectangle. | ||
| """ | ||
|
|
||
| def estimate_transform(self, image_left, image_right, angle_left=0, angle_right=0): | ||
| """Estimate transformation for stitching of images based on 'rectangle' Stitching. | ||
| Args: | ||
| image_left (ndarray): Input left image. | ||
| image_right (ndarray): Input right image. | ||
| angle_left (int): Angle in degree to rotate left image. | ||
| angle_right (int): Angle in degree to rotate right image. | ||
| Returns: | ||
| - **homo_left** (ndarray) -- homography *(3,3)* for ``image_left`` to form a panorama. | ||
| - **homo_right** (ndarray) -- homography *(3,3)* for ``image_right`` to form a panorama. | ||
| - **pano_size** (tuple) -- Size *(width, height)* of the panorama. | ||
| """ | ||
| # TODO(gitmirgut) set all to False | ||
| self.size_left = image_left.shape[:2][::-1] | ||
| self.size_right = image_right.shape[:2][::-1] | ||
|
|
||
| image_left, affine_left = self._prepare_image(image_left, angle_left) | ||
| image_right, affine_right = self._prepare_image(image_right, angle_right) | ||
|
|
||
| rot_size_left = image_left.shape[:2][::-1] | ||
| rot_size_right = image_right.shape[:2][::-1] | ||
| pt_picker = picker.PointPicker() | ||
| pts_left, pts_right = pt_picker.pick([image_left, image_right], False) | ||
| assert len(pts_left) == 4 and len(pts_right) == 4 | ||
| pts_left_srt = helpers.sort_pts(pts_left) | ||
| pts_right_srt = helpers.sort_pts(pts_right) | ||
|
|
||
| target_pts_left = helpers.raw_estimate_rect(pts_left_srt) | ||
| target_pts_right = helpers.raw_estimate_rect(pts_right_srt) | ||
| target_pts_left, target_pts_right = helpers.harmonize_rects( | ||
| target_pts_left, target_pts_right) | ||
|
|
||
| # declare the shift of the right points | ||
| shift_right = np.amax(target_pts_left[:, 0]) | ||
| target_pts_right[:, 0] = target_pts_right[:, 0] + shift_right | ||
| homo_left, __ = cv2.findHomography(pts_left_srt, target_pts_left) | ||
| homo_right, __ = cv2.findHomography(pts_right_srt, target_pts_right) | ||
|
|
||
| # calculate the overall homography including the previous rotation | ||
| homo_left = homo_left.dot(affine_left) | ||
| homo_right = homo_right.dot(affine_right) | ||
|
|
||
| homo_trans, pano_size = helpers.align_to_display_area( | ||
| rot_size_left, rot_size_right, homo_left, homo_right) | ||
|
|
||
| self.homo_left = homo_trans.dot(homo_left) | ||
| self.homo_right = homo_trans.dot(homo_right) | ||
| self.pano_size = pano_size | ||
|
|
||
| # set origin/start and end point for convert measures from px to mm | ||
| self.origin = self.map_left_points(np.array([pts_left_srt[0]])) | ||
| self.end_point = self.map_right_points(np.array([pts_right_srt[1]])) | ||
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.