It's interesting that although this image focusing is possible, and it looks really good, it looks slightly unrealistic. I'm not sure if this is due to my previous experience with photos and seeing a photo with the entire range of people in focus isn't common, or if its because my eyes don't really work like that. Focusing on an entire group of people at once with them being different distances from me is improbable (our eyes basically act as cameras), and thus the last photo (although the focusing looks great!), feels slightly unnatural.
bernardmc8
I wonder if computationally extending the depth of field with the all-focus algorithm is computationally expensive and/or efficient. I feel like in the real world it shouldn’t ultimately end up taking up a lot of computer resources or time because it seems like a common enough operation that would be done but I’m not personally sure.
ananthmrao
@showyouramen, I think this may be because of your photography experience. As a non-photographer, the right-most image looks quite natural to me. Regardless, the degree of predictability in image features that allow such refocusing is amazing! I wonder if it works so well in all cases, though
JefferyYC
I wonder how the refocusing algorithm exactly works. I imagine it should detect the depth of the objects in the photograph in order to fuse multiple focused parts from images with different sensor position.
It's interesting that although this image focusing is possible, and it looks really good, it looks slightly unrealistic. I'm not sure if this is due to my previous experience with photos and seeing a photo with the entire range of people in focus isn't common, or if its because my eyes don't really work like that. Focusing on an entire group of people at once with them being different distances from me is improbable (our eyes basically act as cameras), and thus the last photo (although the focusing looks great!), feels slightly unnatural.
I wonder if computationally extending the depth of field with the all-focus algorithm is computationally expensive and/or efficient. I feel like in the real world it shouldn’t ultimately end up taking up a lot of computer resources or time because it seems like a common enough operation that would be done but I’m not personally sure.
@showyouramen, I think this may be because of your photography experience. As a non-photographer, the right-most image looks quite natural to me. Regardless, the degree of predictability in image features that allow such refocusing is amazing! I wonder if it works so well in all cases, though
I wonder how the refocusing algorithm exactly works. I imagine it should detect the depth of the objects in the photograph in order to fuse multiple focused parts from images with different sensor position.