Why is the result picture more blurred than the original picture? about hdr-plus HOT 5 CLOSED

The algorithm averages patches from different frames in order to decrease noise, however, in addition to decreasing noise, this can soften details (and cause artifacts) in some cases. If you provide the input images and result you are seeing, I could provide further insight.

from hdr-plus.

The following are original image and result @timothybrooks

from hdr-plus.

The pipeline takes a burst of underexposed raw frames, whereas the original image you provided is a processed jpeg. Please provide the raw frames if you have them, otherwise, the pipeline is not expected to work for your application.

from hdr-plus.

Yes you are right, but why does the floor become colorful, which is very different from the original floor. I think this is also the problem. @timothybrooks I also try the underexposed original images, but they change the color more or less after this process. And edges are always unnatural:

from hdr-plus.

There are a few things at play. First, this pipeline is for bursts of underexposed raw images. For it to provide benefit, you will need to use a sequence of raw images as input, not a single image.

Second, there are some weak points in my implementation, notably the demosaic algorithm that I use is not state-of-the-art, and I do not have any chromatic aberration suppression. These two factors, especially on the fisheye lens with strong distortion, causes the issues near edges.

Third, as with all raw image pipelines, color and lighting parameters must be tuned to provide the right look for a given camera. I tuned these variables to my Canon 5D III (and to my personal preferences), but feel free to change the compression factor, amount of contrast, white balance, etc.

I hope this helps, and I encourage you to take a look at my writeup if you have more specific questions about the algorithms I am using. If you have improvements, you are welcome to submit a pull request :)

from hdr-plus.