-
Notifications
You must be signed in to change notification settings - Fork 58
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
confused about img_corrected #23
Comments
Hi, thanks for the question. This is just to normalize the estimation. Note that color constancy has only two DoFs instead of three (see, FAQ (h)). |
If I want to use the ground-truth to correct the picture, img_corrected = np.power(mages[:, :, :, ::-1] / 65535 / illums_ground_truth , 1/2.2), is it right? |
That depends on if you want to adjust the exposure level after color constancy. After you divide the image by pow 1/2.2 is for gamma correction, from linear to sRGB. |
It seems so complex, I'm not clearly understand about that. Now I want to visualize the raw image and corrected image after divided the image by illums_ground_truth , Is there some standard operation to corrected the image ? I just divide the image by illums_ground_truth, but the image is dark. Your mean is that this image is not exposure? So the raw image is not exposure? So visualize the corrected image not only should remove the color cast but need to exposure? |
You are right that color constancy only takes care of the color cast, not the exposure. However, when visualizing them people also want the images to be correctly exposed. Therefore, you do not want |
Thanks for your reply! But I don't understand what image is correctly exposed, is there some baseline? |
A common standard is to scale the RGB values so that their average is 0.18
(and then do the 1/2.2 gamma).
Best,
Yuanming
--------------------------------------------------------------
MIT CSAIL
…On Tue, Aug 21, 2018 at 2:21 AM zyqgmzyq ***@***.***> wrote:
Thanks for your reply! But I don't understand what image is correctly
exposed, is there some baseline?
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#23 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AGP-qFUEJT7NVfP1TbLgZf6F11rlP3fEks5uS6bggaJpZM4V1OhW>
.
|
Sorry,I don't know what RGB value is. Could you please describe it in detail? |
img_corrected = tf.pow(
images[:, :, :, ::-1] / 65535 / illums_pooled[:, None, None, :] *
tf.reduce_mean(illums_pooled, axis=(1), keep_dims=True)[:, None, None, :],
1 / VIS_GAMMA)
Why need to multipy np.mean(illums_pooled)?
The text was updated successfully, but these errors were encountered: