def segment_image(im, idx, scene, depth_threshold=0.1, angle_threshold=20, min_points=100):
floor_idx = [x["idx"] for x in scene if x["model"] == "Floor"][0]
wall_idx = [x["idx"] for x in scene if x["model"] == "Wall"][0]
imc = im.copy()
imc[idx == floor_idx] = 0
imc[idx == wall_idx] = 0
points = compute_points(imc)
normals = compute_normals(points)
return seg.segment_normals(points, normals, depth_threshold, angle_threshold, min_points)
#depth_image = spm.imread('kinect_data/depth.1352980264.888936388.png', True)
idx_image = spm.imread('test_data/3.idx.png')
mask = depth_image == 0
depth_image = cv2.bilateralFilter(depth_image, 9, 30, 100)
depth_image[mask] = 0
pylab.imshow(depth_image)
pylab.show()
points = d.compute_points(depth_image)
normals = d.compute_normals(points)
#normals = seg.compute_normals(points, 0.02, 10)
labels, num_labels = seg.segment_normals(points, normals, 0.1, 10, 100)
#depth_image[labels == 0] = 0
#pylab.imshow(depth_image)
#pylab.show()
max_elements = 4
segments = []
for i in range(num_labels):
v, u = np.nonzero(labels == i)
object_idx = np.unique(idx_image[(v, u)])
if object_idx.shape[0] == 1:
segments.append({'segmentation_label':i, 'object_label':object_idx[0], 'u':u, 'v':v})