Split Image Into Channels
SUMMARY
Split Image Into Channels splits a multi-channel image into its individual channels.
Given an (H, W, C) image, it returns one single-channel (H, W) image per input channel, in the same order as the input (e.g. R, G, B for an RGB image; B, G, R for BGR; R, G, B, A for RGBA). A grayscale (H, W) input is first converted to BGR, so the result is 3 identical channels rather than an error. It is the inverse of merge_image_from_channels — use the two together to process a single channel in isolation and recombine.
Use this Skill when you want to isolate and process a single color channel.
The Skill
from telekinesis import pupil
image_channels = pupil.split_image_into_channels(image=image)
channel_images = image_channels.to_list()Example
Input Image

Original RGB image of a yellow tote holding a carton and a polybag
Channel 1 (Red)

Red channel: the yellow tote is bright
Channel 2 (Green)

Green channel: the yellow tote is also bright
Channel 3 (Blue)

Blue channel: the yellow tote turns dark, separating it from the white label
The Code
"""Demonstrates splitting an image into color channels."""
from loguru import logger
import rerun as rr
from telekinesis import pupil, datatypes
def split_image_into_channels_example():
"""Splits an image into its color channels."""
# ===================== Load Image ==========================================
image_url = "https://assets.telekinesis.ai/examples/v1/images/armbench_yellow_tote_carton_polybag.jpg"
image = datatypes.Image.from_url(image_url)
# ===================== Run Skill ==========================================
image_channels = pupil.split_image_into_channels(image=image)
# ===================== Log ================================================
logger.success(f"Split {image} into channels")
logger.success(f"Result: {image_channels}")
# ===================== Visualization (Optional) ======================
rr.init("split_image_into_channels_example", spawn=True)
datatypes.visualize(image, entity_path="1-Original")
channel_names = ["Red", "Green", "Blue"]
for i, channel_image in enumerate(image_channels):
datatypes.visualize(channel_image, entity_path=f"{i + 2}-{channel_names[i]}")
if __name__ == "__main__":
split_image_into_channels_example()Runnable examples are available in the Telekinesis examples repository.
Follow the README in that repository to set up the environment, run this specific example with:
cd telekinesis-examples
python examples/image_processing/split_image_into_channels.pyParameter Configuration
| Parameter | Type | Default | Description |
|---|---|---|---|
image | datatypes.Image | np.ndarray | required | The input image to split, shape (H, W, C). A grayscale (H, W) image is converted to BGR first, producing 3 identical channels |
Returns
| Type | Description |
|---|---|
datatypes.ImageBatch | One single-channel datatypes.Image per input channel, in input-channel order. Call .to_list() for a plain list[datatypes.Image], or index/iterate the batch directly |
Raises
| Exception | Condition |
|---|---|
TypeError | image has an invalid type |
ConfigurationError | The TELEKINESIS_API_KEY environment variable is not set |
SerializationError | The request input failed to serialize, the response was not returned as an Arrow stream, or the response failed to deserialize |
RequestTimeoutError | The request to the Pupil service timed out |
TransportError | A network failure occurred before a response was received |
ClientError | The Pupil service rejected the request due to invalid or malformed input (HTTP 400/422), an unrecognized endpoint (HTTP 404), or another unexpected 4xx response |
AuthenticationError | The API key was rejected as invalid or expired (HTTP 401) |
AuthenticationServiceError | The authentication service returned an invalid response, was temporarily unavailable, or timed out (HTTP 502/503/504) |
ServerError | The Pupil service returned a 5xx or otherwise unexpected error response |
How to Tune the Parameters
The split_image_into_channels Skill takes a single input and has no tunable behavior — the number and order of output channels is fully determined by the input image.
image
- Controls: Which image is split, and therefore how many channels come out (3 for RGB/BGR, 4 for RGBA, 3 identical channels for a grayscale input) and in what order.
- Default: required, no default
- No numeric tuning is available; the output order always mirrors the input's channel order.
TIP
Best practice: Know your input's channel order before indexing the result — index 0 is R for an RGB image but B for a BGR image. Convert with convert_image_color_space first if you need a specific, known order.
Where to Use the Skill
Common pipelines include:
- Per-channel processing – Apply a filter or enhancement to a single channel (e.g. blur only the blue channel) before recombining with
merge_image_from_channels - Channel-based analysis – Inspect or threshold one channel's intensity distribution independently
- Channel swapping – Reorder channels (e.g. RGB to BGR) by splitting and merging in a different order
Alternative Skills
| Skill | vs. Split Image Into Channels |
|---|---|
| merge_image_from_channels | The inverse operation — recombines single-channel images into one multi-channel image. Use it after processing the channels this Skill produces. |
| convert_image_color_space | Converts the whole image to a different color space in one call. Use this Skill instead only when you need per-channel access rather than a full conversion. |
When Not to Use the Skill
Do not use Split Image Into Channels when:
- You only need a single channel, not all of them (index the array directly, or convert to grayscale with
convert_image_color_space) - The image is already single-channel (grayscale) (splitting just duplicates it three times — a no-op you can skip)
- You need channels of a different color space (e.g. HSV, LAB) (convert with
convert_image_color_spacefirst, then split)