YUY2 format to rgb format

[waiyc]

Hi,

Is there anyways to convert the YUY2 format from RGB camera to rgb format using opencv.

config.enable_stream(rs.stream.color, 1920, 1080, rs.format.yuyv, 15)
...
color_frame = frames.get_color_frame()
yuyv_image = np.asanyarray(color_frame.get_data())
rgb_image = cv2.cvtColor(yuyv_image, cv2.COLOR_YUV2BGR_YUY2)

I have tried using cv2,cvtColor() with COLOR_YUV2BGR_YUY2 but failed to convert.

error: (-215) scn == 2 && depth == 0 in function cvtColor

[MartyG-RealSense]

I wonder if you might be able to adapt the conversion information in the link below.

https://answers.opencv.org/question/87083/yuyv422-to-bgr/?answer=202592#post-id-202592

[waiyc]

@MartyG-RealSense thanks for the reply.
I have tried the same ColorConversionCodes mentioned in the discussion, but still facing the same error.

[MartyG-RealSense]

Apologies for the delay in responding further. I was carefully researching your error message. I came across this case with the same error:

https://stackanswers.net/questions/error-trying-to-use-cvtcolor-with-cv2-color-yuv2bgr-y422-error-215-scn-2-depth-0-in-function-cv-cvtcolor

From what I can gather (I am not an OpenCV expert), the scn==2 error may occur because the convert function needs an image with 2 channels.

[waiyc]

yes, in order to convert YUY2 to rgb the source image is required to contains 2 channels.

However, the input array size after reading the rgb image with input YUYV format contains a single channel 2d array only.

config.enable_stream(rs.stream.color, D435_RAW_COLOR_W, D435_RAW_COLOR_H, rs.format.yuyv, D435_RAW_COLOR_FPS)
color_frame = frames.get_color_frame()
yuyv_image = np.asanyarray(color_frame.get_data())

yuyv_image array size = (1080, 1920)
yuyv_image array data type = uint16

---

i also came across show_yuv_frame() developed few years ago shows that cv2.cvtColor(frame, cv2.COLOR_YUV2RGB_YUYV) can be used to convert YUY2 format to rgb without issues previously and the color_frame array should have 2 channels.

[RealSenseCustomerSupport]

---

@waiyc Sorry for later response. Did you get it through now? Looking forward to your update. Thanks!

[RealSenseCustomerSupport]

---

@waiyc Any update? Thanks!

[RealSenseCustomerSupport]

---

@waiyc Will close the issue if no other questions. Thanks!

[tschoepflin]

Here's a solution... the trick is that YUY2 format needs to be put into Y (channel 0) and UV (channel 1). Moving forward I will never use YUY2 with OpenCV since it is heavily biased toward BGR.

# Read BAG file and store image

import pyrealsense2 as rs
import numpy as np
import cv2
import sys, getopt

def GetBGR(frame_color):
    # Input: Intel handle to 16-bit YU/YV data
    # Output: BGR8
    H = frame_color.get_height()
    W = frame_color.get_width()
    Y = np.frombuffer(frame_color.get_data(), dtype=np.uint8)[0::2].reshape(H,W)
    UV = np.frombuffer(frame_color.get_data(), dtype=np.uint8)[1::2].reshape(H,W)
    YUV =  np.zeros((H,W,2), 'uint8')
    YUV[:,:,0] = Y
    YUV[:,:,1] = UV
    BGR = cv2.cvtColor(YUV,cv2.COLOR_YUV2BGR_YUYV)
    return BGR


# Configure depth and color streams
config = rs.config()
pipeline = rs.pipeline()
rs.config.enable_device_from_file(config, 'g_cam0.bag')
config.enable_stream(rs.stream.color, format=rs.format.yuyv, framerate=15)

profile = pipeline.start(config)

try:
    # Get 5 images
    for i in range(60):
        frames = pipeline.wait_for_frames()
        frame_color = frames.get_color_frame()
        if frame_color:
            BGR = GetBGR(frame_color)
            cv2.imwrite(f'RS_RGB_img_{i}.jpg', BGR)

finally:
    # Stop streaming
    pipeline.stop()

[MartyG-RealSense]

Thanks so much for sharing your solution with the RealSense community @tschoepflin :)