reproject_rgbd#

deltacamera.reproject_rgbd(image, depth_map, old_camera, new_camera, output_imshape=None, border_mode=cv2.BORDER_CONSTANT, border_value=0, image_interp=None, depth_interp=cv2.INTER_NEAREST, antialias_factor=1, dst=None, cache_maps=False, precomp_undist_maps=True, use_linear_srgb=False, return_validity_mask=False)[source]#

Reproject an RGB image and depth map together, sharing intermediate computation.

This is equivalent to calling reproject_image and reproject_depth_map separately, but avoids redundant computation of remap maps, homographies, and validity masks.

When antialias_factor > 1, both are rendered at higher resolution and then block-downsampled: the image via area averaging, the depth via nanmedian (NaN-safe).

Parameters:

  • image (ndarray) – RGB image, shape (H, W, 3).

  • depth_map (ndarray) – Depth map, shape (H, W), float. NaN or 0 marks invalid pixels.

  • old_camera (Camera) – The camera that captured the image and depth.

  • new_camera (Camera) – The target camera.

  • output_imshape (tuple) – (height, width) for the output. If None, uses new_camera.image_shape.

  • border_mode (int) – OpenCV border mode for treating pixels outside the image.

  • border_value – Border value for image pixels outside the input.

  • image_interp – OpenCV interpolation for the image. Default: cv2.INTER_LINEAR.

  • depth_interp (int) – OpenCV interpolation for the depth. Default: cv2.INTER_NEAREST.

  • antialias_factor (int) – Supersample factor. Both image and depth are rendered at this multiple of the output resolution, then downsampled (image: area average, depth: nanmedian).

  • dst – Destination array for the image (optional).

  • cache_maps (bool) – Whether to cache coordinate maps for repeated use.

  • precomp_undist_maps (bool) – Whether to precompute and cache undistortion maps.

  • use_linear_srgb (bool) – If True, decode image from 8-bit sRGB to 16-bit linear space before reprojecting and encode back afterward.

  • return_validity_mask (bool) – If True, return (new_image, new_depth, validity_mask).

Returns:

(new_image, new_depth) or (new_image, new_depth, validity_mask).

Return type:

tuple