Ce Liu, Jenny Yuen, Antonio Torralba,JosefSivic, andWilliam T. Freeman 版权所有。
修改的部分函数与变量的名字，使其好懂了一些。不过这个算法由于没有salient keypoint detection和rotation normalization，所以对尺度以及旋转这些affine transform没有移不变的性质。
function [ SIFTFeatureVector, locationX, locationY ] = DenseSIFT( image, nPatchSize, nGridSpacing ) image = double( image ); image = mean( image, 3 ); image = image / max( image( : ) ); % parameters nAngleNums = 8; nBinNums = 4; nSampleNums = nBinNums * nBinNums; alpha = 9; %% parameter for attenuation of angles (must be odd) if nargin < 5 sigmaGuassian = 1; end angleStep = 2 * pi / nAngleNums; angles = 0 : angleStep : 2 * pi; angles( nAngleNums 1 ) = [ ]; % bin centers [ nRow nCol ] = size( image ); [ gaus