imread

Imread is a function used to load an image from a file into in-memory data, producing an array or image object that can be processed by other image processing routines. The exact return type, color ordering, and additional options vary between libraries, but the general purpose is to convert a file into a pixel array for manipulation.

In OpenCV, the function is cv2.imread(filename, flags) and returns a NumPy array with shape (height, width, channels). Flags include IMREAD_COLOR (default), IMREAD_GRAYSCALE, and IMREAD_UNCHANGED. Color channels in OpenCV are ordered BGR, not RGB. If the image cannot be read, cv2.imread returns None. To display or work with standard RGB data, a conversion to RGB may be necessary, e.g., cv2.cvtColor(img, cv2.COLOR_BGR2RGB).

In libraries such as matplotlib and scikit-image, imread returns a NumPy array with pixel values typically in the image’s native range and dtype. JPEG images commonly yield uint8 arrays in the range [0, 255], while PNGs with transparency may include an alpha channel. Some libraries normalize data to [0, 1] for floating-point processing or for display.

In PIL (Pillow), the typical workflow uses Image.open to load an image and then convert to a NumPy array if needed (numpy.asarray) or to a specific mode (e.g., convert('RGB')) for consistent channel ordering.

Considerations when using imread include handling missing or unsupported files, memory usage for large images, and the need for color space or channel conversions to ensure compatibility with downstream processing or visualization workflows.