何凯明发表论文

2003年博士毕业于西安交通大学人工智能与机器人研究所，此后孙剑一直在微软亚洲研究院工作，担任首席研究员，其主要研究方向是计算摄影学，人脸识别和基于深度学习的图像理解。他拥有超过40项专利，自2002年以来在顶级学术会议和期刊上发表学术论文100余篇，被引数为255630。并于2009和2016年两次获得CVPR计算机视觉年会的最佳论文奖。孙剑博士是一位在人工智能技术领域探索和创新的领路人。

%%%%暗原色去雾算法是建立在户外自然场景暗通道优先法则的基础上的去雾方法，其实就是解一个方程%%I(x)=J(x)t(x)+A(1-t(x));其中I(x)是受到雾气污染的图像。J(x)是我们需要求的去雾后的图像%%t(x)是天空中云层的透射分布率，A是天空的亮度%暗影去雾算法% 原始图像img_name = imread('C:\Users\Administrator\Desktop\桌面文件\专业课\设计\2.bmp');I = double(img_name) / 255;% 获取图像大小[h,w,c] = size(I);%去雾系数w0 = 0.95;img_size = w * h;%初始化结果图像dehaze = zeros(h,w,c);%初始化暗影通道图像win_dark = zeros(h,w);for i=1:h for j=1:wwin_dark(i,j)=min(I(i,j,:));%将三个通道中最暗的值赋给dark_I(i,j),显然，三维图变成了二维图endendwin_dark = ordfilt2(win_dark,1,ones(9,9),'symmetric');%计算大气亮度A，相关原理详见论文“Single Image Haze Removal Using Dark Channel Prior”dark_channel = win_dark;A = max(max(dark_channel));[i,j] = find(dark_channel==A);i = i(1);j = j(1);A = mean(I(i,j,:));%计算初始的transmission maptransmission = 1 - w0 * win_dark / A;%用guided filter对trasmission map做soft mattinggray_I = I(:,:,1);%这里gray_I 可以是RGB图像中任何一个通道p = transmission;r = 80;eps = 10^-3;transmission_filter = guidedfilter(gray_I, p, r, eps);t0=0.1;t1 = max(t0,transmission_filter);for i=1:cfor j=1:hfor l=1:wdehaze(j,l,i)=(I(j,l,i)-A)/t1(j,l)+A;endendendfigure,imshow(I);title('去雾前')figure,imshow(dehaze);title('去雾后')%里面用到的两个函数如下：（这两个函数是何凯明guided filter中的，详细原理可以参考博客园中laviewpb的分析）function q = guidedfilter(I, p, r, eps)% GUIDEDFILTER O(1) time implementation of guided filter.%% - guidance image: I (should be a gray-scale/single channel image)% - filtering input image: p (should be a gray-scale/single channel image)% - local window radius: r% - regularization parameter: eps[hei, wid] = size(I);N = boxfilter(ones(hei, wid), r); % the size of each local patch; N=(2r+1)^2 except for boundary pixels.mean_I = boxfilter(I, r) ./ N;mean_p = boxfilter(p, r) ./ N;mean_Ip = boxfilter(I.*p, r) ./ N;cov_Ip = mean_Ip - mean_I .* mean_p; % this is the covariance of (I, p) in each local patch.mean_II = boxfilter(I.*I, r) ./ N;var_I = mean_II - mean_I .* mean_I;a = cov_Ip ./ (var_I + eps); % Eqn. (5) in the paper;b = mean_p - a .* mean_I; % Eqn. (6) in the paper;mean_a = boxfilter(a, r) ./ N;mean_b = boxfilter(b, r) ./ N;q = mean_a .* I + mean_b; % Eqn. (8) in the paper;endfunction imDst = boxfilter(imSrc, r)% BOXFILTER O(1) time box filtering using cumulative sum%% - Definition imDst(x, y)=sum(sum(imSrc(x-r:x+r,y-r:y+r)));% - Running time independent of r; % - Equivalent to the function: colfilt(imSrc, [2*r+1, 2*r+1], 'sliding', @sum);% - But much faster.[hei, wid] = size(imSrc);imDst = zeros(size(imSrc));%cumulative sum over Y axisimCum = cumsum(imSrc, 1);%difference over Y axisimDst(1:r+1, :) = imCum(1+r:2*r+1, :);imDst(r+2:hei-r, :) = imCum(2*r+2:hei, :) - imCum(1:hei-2*r-1, :);imDst(hei-r+1:hei, :) = repmat(imCum(hei, :), [r, 1]) - imCum(hei-2*r:hei-r-1, :);%cumulative sum over X axisimCum = cumsum(imDst, 2);%difference over Y axisimDst(:, 1:r+1) = imCum(:, 1+r:2*r+1);imDst(:, r+2:wid-r) = imCum(:, 2*r+2:wid) - imCum(:, 1:wid-2*r-1);imDst(:, wid-r+1:wid) = repmat(imCum(:, wid), [1, r]) - imCum(:, wid-2*r:wid-r-1);end

function dark = darkChannel(imRGB)r=imRGB(:,:,1);g=imRGB(:,:,2);b=imRGB(:,:,3); 分别提取三色的灰度图[m n] = size(r); 提取单色图矩阵的宽度和长度a = zeros(m,n); 创建m*n的零矩阵afor i = 1: mfor j = 1: na(i,j) = min(r(i,j), g(i,j)); a(i,j)= min(a(i,j), b(i,j)); 依次比较三色分量的最小值提取为暗通道图endendd = ones(15,15); 创建15*15的单位矩阵fun = @(block_struct)min(min(block_struct.data))*d; dark = blockproc(a, [15 15], fun); 将图片分成15*15的小块并将每一块变成其中的最小值dark = dark(1:m, 1:n);我也是新手啊兄弟只能帮你到这儿了

孙剑博士一生专注于科研工作。他的不幸离世，让旷视失去了一位在人工智能技术领域探索和创新的领路人。每一位和他共事过的旷视同学，失去了一位智慧谦和的师长。旷视研究院的墙上一直挂着孙剑博士奠定的研究使命：持续创新拓展认知边界，非凡科技成就产品价值。这句话将永远激励我们，勇攀科研高峰！