Displaying Images¶

This page describes how to display images on the Coral Dev Board's HDMI output. Four methods — GStreamer, ImageMagick, OpenCV, and Pillow — were tested on the actual board, and all formats (BMP, JPG, PNG, WebP) were confirmed to work with each method.

Display Environment¶

The display stack on Coral Dev Board (Mendel Linux) is configured as follows:

Weston (Wayland compositor) runs by default
XWayland is enabled — X11 applications work out of the box
/dev/fb0 exists but the i.MX 8M uses DRM/KMS, so direct framebuffer access requires stopping Weston

Method 1: GStreamer (Recommended)¶

Displays the image as a Wayland window. No need to stop Weston.

Installation¶

sudo apt-get install gstreamer1.0-tools

Command¶

WAYLAND_DISPLAY=wayland-0 XDG_RUNTIME_DIR=/run/user/1000 \
  gst-launch-1.0 filesrc location=<file> ! decodebin ! videoconvert ! imagefreeze ! waylandsink

decodebin automatically detects the format and decodes it, so the same command works for BMP, JPG, PNG, and WebP.

Screen not refreshing after exit

The screen may not update after GStreamer exits. Run sudo systemctl restart weston to recover.

Method 2: ImageMagick¶

Displays via XWayland. No need to stop Weston. Provides GUI controls (zoom, rotate, color adjustment, etc.) with the shortest command.

Installation¶

sudo apt-get install imagemagick

Command¶

DISPLAY=:0 display -geometry +0+0 <file>

Without -geometry +0+0, the window may be placed at the bottom of the screen, causing the image to be clipped.

Method 3: OpenCV (python3-opencv)¶

Displays via XWayland. No need to stop Weston. Useful for combining inference result rendering (rectangles, text, etc.) and display in a single library.

Installation¶

sudo apt-get install python3-opencv

pip install not supported

pip install opencv-python-headless fails due to numpy build errors. Always install via apt.

Code Example¶

import cv2
import os

os.environ["WAYLAND_DISPLAY"] = "wayland-0"
os.environ["XDG_RUNTIME_DIR"] = "/run/user/1000"

img = cv2.imread("<file>")
cv2.imshow("viewer", img)
cv2.waitKey(0)
cv2.destroyAllWindows()

Method 4: Pillow + GStreamer¶

Load the image with Pillow, convert to raw bytes, and pipe into GStreamer's rawvideoparse. Pillow is pre-installed so no additional installation is required, but GStreamer is needed.

Since GStreamer's decodebin can do the same thing on its own, this method is only useful when you need Python-based image processing before display.

Comparison¶

Method	Package	Stop Weston	Notes
GStreamer	`gstreamer1.0-tools`	No	CLI-only, flexible pipelines
ImageMagick	`imagemagick`	No	GUI controls, shortest command
OpenCV	`python3-opencv` (apt)	No	Easy to integrate into Python inference pipelines
Pillow + GStreamer	`gstreamer1.0-tools`	No	Display after Python image processing

Supported Formats¶

All tools were verified on the actual board with BMP / JPG / PNG / WebP.