normalization.py

import numpy as np
from .bayer_chan_mixer import bayer_to_rgbg, rgbg_to_bayer

def bayer_normalize(rgbg : np.ndarray, chan_black : np.ndarray, chan_sat : np.ndarray) -> np.ndarray:
    """Normalize a Bayer image from sensor range to [0,1].

    This converts the input to float32 and applies the black floor and saturation clipping.

    Args:
        rgbg (np.ndarray): Raw Bayer image in RGBG format.
        chan_black (np.ndarray): Black level for each channel (minimum length 4).
        chan_sat (np.ndarray): Saturation level for each channel (minimum length 4).

    Returns:
        np.ndarray: Normalized Bayer image.
    """
    r, g1, b, g2 = bayer_to_rgbg(rgbg)

    # TODO - Check channel ordering, should be in Bayer pattern
    r   = np.clip(r - chan_black[0], 0, chan_sat[0]).astype(np.float32) / chan_sat[0]
    g1  = np.clip(g1 - chan_black[1], 0, chan_sat[1]).astype(np.float32) / chan_sat[1]
    b   = np.clip(b - chan_black[2], 0, chan_sat[2]).astype(np.float32) / chan_sat[2]
    g2  = np.clip(g2 - chan_black[3], 0, chan_sat[3]).astype(np.float32) / chan_sat[3]

    return rgbg_to_bayer(r, g1, b, g2)