Use to see how much image information is preserved in filter outputs. Reconstructs the orginal image from filter outputs (approximately). The filter output for a patch is given by IFR=F*P where F is the set of filters in matrix form, P is the patch and IFR is the filter responses at the center of the patch. We want to recover P from IFR and F, this is underconstrained but a solution can be found using least squared. Note that each recovered P will be 0 mean if no mean information is captured by the filter outputs. Can apply to a single patch (interatctively specified), or the entire image (by keeping the central pixel from each patch). USAGE I2 = FbReconstruct2d( I, FB, patch ) INPUTS I - original image FB - FB to apply and do reconstruction with patch - reconstruct just patch or entire image OUTPUTS I2 - reconstructed image / patch EXAMPLE load trees; X=imresize(X,.5); load FbDoG.mat; I2 = FbReconstruct2d( X, FB, 0 ); See also FBAPPLY2D Piotr's Computer Vision Matlab Toolbox Version 2.0 Copyright 2014 Piotr Dollar. [pdollar-at-gmail.com] Licensed under the Simplified BSD License [see external/bsd.txt]