shojtsy Posted January 15, 2018 Posted January 15, 2018 This thread gave me an idea to create an estimated Depth mask: I have created a macro to recover some depth information from the Remove Haze filter itself. Create_depth_mask.afmacro The way you use this is to select your pixel layer, run the macro, and then drag the resulting layer group into a mask position of some adjustment. The red tint is intentional to help previewing the mask. When the layer group is used as a mask of some other layer, the tint will be ignored, and only the alpha channel will be used. See example usage with a black fill layer being masked by the depth mask. Create_depth_mask.afmacro dmstraker 1 Quote
BornInPastCentury Posted January 15, 2018 Posted January 15, 2018 Hi. Nice and very informative (with help of little reverse engineering ). However, when this macro is executed for image with little to no haze, the resultant mask seems to not show the depth... Result may indicate that the macro (and dehaze filter itself) estimates amount of haze rather than the scene depth. Quote
shojtsy Posted January 15, 2018 Author Posted January 15, 2018 6 minutes ago, BornInPastCentury said: Hi. Nice and very informative (with help of little reverse engineering ). However, when this macro is executed for image with little to no haze, the resultant mask seems to not show the depth... Result may indicate that the macro (and dehaze filter itself) estimates amount of haze rather than the scene depth. Maybe you are looking for defocus estimation? Or some artificial intelligence that understands the objects? I don't think either is possible in Affinty. Quote
BornInPastCentury Posted January 16, 2018 Posted January 16, 2018 23 hours ago, shojtsy said: Maybe you are looking for defocus estimation? Or some artificial intelligence that understands the objects? I don't think either is possible in Affinty. There are numerous publications on the depth estimation subject. Many uses convolutional neural networks, which is machine learning class algorithm rather than AI. Whatever "engine" uses our Affinity, I think that utilization of KNN is not impossible. Quote
shojtsy Posted January 17, 2018 Author Posted January 17, 2018 Both single-image defocus estimation and neural networks are possible solutions for this. I am just saying that Affinity doesn't provide the tools for us to do either, and I think also doesn't implement either of them itself. Remove Haze filter probably uses some simple heuristic of detecting grayish gradients and make them contrasty. Quote
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.