(1) Transposition: If any rotation has been requested, the input is transposed to rotate the image (rotation is always implemented by some combination of transposition and flips). (2) Black level compensation: Use the non-light sensing elements (typically in a covered border) to determine what level of charge represents “optically black”. (3) Lens shading: The camera firmware includes a table that corrects for chromatic distortion from the standard module’s lens. This is one reason why third party modules incorporating different lenses may show non-uniform color across a frame. (4) White balance: The red and blue gains are applied to correct the color balance. See awb_gains and awb_mode. (5) Digital gain: As mentioned above, this is a straight-forward post-processing step that applies a gain to the Bayer values. See digital_gain. (6) Bayer de-noise: This is a noise reduction algorithm run on the frame data while it is still in Bayer format. (7) De-mosaic: The frame data is converted from Bayer format to YUV420 which is the format used by the remainder of the pipeline. (8) YUV de-noise: Another noise reduction algorithm, this time with the frame in YUV420 format. See image_denoise and video_denoise. (9) Sharpening: An algorithm to enhance edges in the image. See sharpness. (10) Color processing: The brightness, contrast, and saturation adjustments are implemented. (11) Distortion: The distortion introduced by the camera’s lens is corrected. At present this stage does nothing as the stock lens isn’t a fish-eye lens; it exists as an option should a future sensor require it. (12) Resizing: At this point, the frame is resized to the requested output resolution (all prior stages have been performed on “full” frame data at whatever resolution the sensor is configured to produce). See resolution.