ColorScope

Tool for analyze the image quality

Requirements

• python-opencv
• matplotlib
• scikit-image

$ sudo pip install opencv-python matplotlib scikit-image

Usage

Output format supported: rgb, yuv, hsv, hls

$ ./colorscope.py -i image.jpeg -out_fmt=rgb
R      G      B
23     24     232
255    255    255
...

Raw input images

$ ./colorscope.py -i image.yuv -pix_fmt=nv21 -s 640x480 -out_fmt=rgb

Measure and plot data

$ ./colorscope.py -i reference.jpeg -out_fmt=hls -o ref.json
$ ./colorscope.py -i capture.jpeg -out_fmt=hls -o cap.json
$ ./colorscope.py -gen ref.json cap.json

Quality metrics

Compare of two images quality using PSNR and SSIM metric for multichannel

$ ./colorscope.py -cp metrics reference_image_dir ref_pixel_format ref_video_size capture_image_dir cap_pixel_format cap_video_size

For non raw images video size and pixel format can be ommited

$ ./colorscope.py -scp 0 ssim reference.jpg capture.jpg
$ ./colorscope.py -scp 0 psnr reference.jpg capture.jpg

Multichannel examples:

$ ./colorscope.py -cp ssim  reference.yuv nv12 1920x1080 capture.yuv nv12 1920x1080
$ ./colorscope.py -cp ssim reference.jpg capture.jpg
$ ./colorscope.py -cp psnr reference.jpg capture.jpg

Compare of two images quality using PSNR and SSIM metric for single channel

$ ./colorscope.py -scp metrics channel_number reference_image_dir ref__pxl_format ref_video_size capture_image_dir cap_pxl_format cap_video_size

Channel should be given accordingly to openCV color representation: For BGR (typical way openCV stores RGB)

• blue 0
• green 1
• red 2

For YUV:

• Y 0
• U 1
• V 2

Single channel examples

$ ./colorscope.py -scp ssim 0  reference.yuv nv12 1920x1080 capture.yuv nv12 1920x1080
$ ./colorscope.py -scp psnr 2 reference.jpg capture.jpg