基于SVM的图像无参考质量评价系统matlab实现+仿真录像

function feat = brisque_feature(imdist) %------------------------------------------------ % Feature Computation %------------------------------------------------- scalenum = 2; window = fspecial('gaussian',7,7/6); window = window/sum(sum(window)); feat = []; tic for itr_scale = 1:scalenum mu = filter2(window, imdist, 'same'); mu_sq = mu.*mu; sigma = sqrt(abs(filter2(window, imdist.*imdist, 'same') - mu_sq)); structdis = (imdist-mu)./(sigma+1); [alpha overallstd] = estimateggdparam(structdis(:)); feat = [feat alpha overallstd^2]; shifts = [ 0 1;1 0 ; 1 1; -1 1]; for itr_shift =1:4 shifted_structdis = circshift(structdis,shifts(itr_shift,:)); pair = structdis(:).*shifted_structdis(:); [alpha leftstd rightstd] = estimateaggdparam(pair); const =(sqrt(gamma(1/alpha))/sqrt(gamma(3/alpha))); meanparam =(rightstd-leftstd)*(gamma(2/alpha)/gamma(1/alpha))*const; feat =[feat alpha meanparam leftstd^2 rightstd^2]; end imdist = imresize(imdist,0.5); end toc