Is there code of preprocessing data? about dped HOT 4 CLOSED

Hi,

A detailed description of the patch extraction algorithm is provided in the paper. Mainly, in consists of two parts:

1. Rough image alignment using SIFT features (you can use this code as a reference: http://www.vlfeat.org/install-matlab.html)

2. Extracting precisely aligned patches from pre-aligned images obtained on the previous step using a sliding window approach. Basically, two sliding windows are shifting in parallel along a DSLR-phone image pair, and the position of one window is additionally adjusted by small shifts and rotations to maximize cross-entropy between the observed patches. If this cross-entropy is larger than a threshold (0.9), two current patches (from DSLR and phone image) are added into the training dataset.

Hope this helps.

from dped.

Is there any detail instruction about extracting patches @aiff22? Thank you

from dped.

Hi,

A detailed description of the patch extraction algorithm is provided in the paper. Mainly, in consists of two parts:

1. Rough image alignment using SIFT features (you can use this code as a reference: http://www.vlfeat.org/install-matlab.html)
2. Extracting precisely aligned patches from pre-aligned images obtained on the previous step using a sliding window approach. Basically, two sliding windows are shifting in parallel along a DSLR-phone image pair, and the position of one window is additionally adjusted by small shifts and rotations to maximize cross-entropy between the observed patches. If this cross-entropy is larger than a threshold (0.9), two current patches (from DSLR and phone image) are added into the training dataset.

Hope this helps.

Hi, aiff22,
You mentioned that the position of one window is additionally adjusted by small shifts and rotations. However, I still don't understand how to do it. Can you specify the details?

from dped.

2. cross-entropy

Can you provide the Extracting code ?

from dped.