基于塊的全搜索運動估計算法實現(xiàn)實驗報告(共10頁)

1、精選優(yōu)質(zhì)文檔-傾情為你奉上數(shù)字視頻處理實驗報告學 院： 通信與信息工程學院系 班： 電信科0901班姓 名： 學 號： 時 間： 2012 年11月23號一、實驗名稱：基于塊的全搜索運動估計算法實現(xiàn)二、實驗?zāi)康模?、掌握運動估計算法的實現(xiàn)原理。2、掌握運動估計算法的研究現(xiàn)狀及多種計算方法。3、學習基于塊的全搜索運動估計算法，研究分析其Matlab實現(xiàn)程序過程，并補充完成程序，對實驗結(jié)果進行分析比較。 三、實驗要求三、實驗要求1、 對實驗程序motionEstAnalysis.m進行分析，完成主程序流程圖。函數(shù)流程圖：2、編寫補充完成部分不全程序代碼，調(diào)試程序使其能正確運行(1) motionE

2、stES( )% Computes motion vectors using exhaustive search method（全搜索法計算運動矢量）% Input% imgP : The image for which we want to find motion vectors（當前圖像）% imgI : The reference image（參考圖像）% mbSize : Size of the macroblock（宏塊尺寸）% p : Search parameter (read literature to find what this means)（搜索參數(shù)）% Ouput% m

3、otionVect : the motion vectors for each integral macroblock in imgP（當前圖像中每一個積分宏塊的運動矢量）% EScomputations: The average number of points searched for a macroblock（每個宏塊搜索的平均點數(shù)）% Written by Aroh Barjatyafunction BlockCenter, motionVect, EScomputations = motionEstES(imgP, imgI, mbSize, p) % 定義函數(shù)文件motionEst

4、ES.m，imgP、 imgI、 mbSize、 p為傳入?yún)?shù)，BlockCenter、motionVect、 EScomputations為返回參數(shù)row col = size(imgI); % 將參考圖像的行數(shù)賦值給row，列數(shù)賦值給colblockcenter = zeros(2,row*col/mbSize2);vectors = zeros(2,row*col/mbSize2); % 定義全0的矢量矩陣的大小costs = ones(2*p + 1, 2*p +1) * 65537; % 定義最小絕對差矩陣的大小computations = 0; % 搜索點數(shù)賦初值為0 % we s

5、tart off from the top left of the image（從圖像左上角開始）% we will walk in steps of mbSize（以宏塊尺寸為步長）% for every marcoblock that we look at we will look for% a close match p pixels on the left, right, top and bottom of it （對于每一個宏塊，在它的上下左右找到與搜索參數(shù)p最匹配的像素）mbCount = 1; %搜索的宏塊數(shù)賦初值為1%1為循環(huán)起始值，mbSize為步長值，row-mbSize+

6、1為循環(huán)終止值for i = 1 : mbSize : row-mbSize+1 for j = 1 : mbSize : col-mbSize+1 % the exhaustive search starts here（全搜索開始） % we will evaluate cost for (2p + 1) blocks vertically % and (2p + 1) blocks horizontaly（我們將計算水平方向上(2p + 1)個塊的最小絕對差和垂直方向上(2p + 1)個塊的最小絕對差） % m is row(vertical) index（m為行指數(shù)） % n is co

7、l(horizontal) index（n為列指數(shù)） % this means we are scanning in raster order for m = -p : p %水平方向上位移矢量范圍 for n = -p : p %垂直方向上位移矢量范圍 % 補充下面程序 % row/Vert co-ordinate for ref block （參考塊的行(垂直方向)的范圍） refBlkVer = i+m; % col/Horizontal co-ordinate（參考塊的列(水平方向)的范圍） refBlkHor = j+n; %如果參考塊的行列范圍的任意一個在已經(jīng)搜索過的宏塊之外，則繼

8、續(xù)下一步的搜索 if ( refBlkVer < 1 | refBlkVer+mbSize-1 > row . | refBlkHor < 1 | refBlkHor+mbSize-1 > col) continue; end costs(m+p+1,n+p+1) = costFuncMAD(imgP(i:i+mbSize-1,j:j+mbSize-1), . imgI(refBlkVer:refBlkVer+mbSize-1, refBlkHor:refBlkHor+mbSize-1), mbSize); % 搜索下一個點 computations = computa

9、tions + 1; end end % Now we find the vector where the cost is minimum % and store it . this is what will be passed back.（現(xiàn)在找到有最小絕對差的矢量并存儲它，這就是將被返回的東西） % 補充下面程序 blockcenter(1,mbCount) = i+ mbSize/2-1; blockcenter(2,mbCount) = j+ mbSize/2-1; % finds which macroblock in imgI gave us min Cost（找到參考圖像中最小絕

10、對差的宏塊） dx, dy, min = minCost(costs); % row co-ordinate for the vector（矢量的行集合） vectors(1,mbCount) = dy-p-1; % col co-ordinate for the vector（矢量的列集合） vectors(2,mbCount) = dx-p-1; %搜索下一個宏塊 mbCount = mbCount + 1; costs = ones(2*p + 1, 2*p +1) * 65537; endendBlockCenter = blockcenter;motionVect = vectors

11、; %返回當前圖像中每一個積分宏塊的運動矢量EScomputations = computations/(mbCount - 1); %返回每個宏塊搜索的平均點數(shù) (2) costFuncMAD( )% Computes the Mean Absolute Difference (MAD) for the given two blocks（對給定的兩個塊計算最小絕對差）% Input% currentBlk : The block for which we are finding the MAD（當前塊）% refBlk : the block w.r.t. which the MAD is

12、being computed（參考塊）% n : the side of the two square blocks% Output% cost : The MAD for the two blocks（兩個塊的最小絕對差）% Written by Aroh Barjatya% 定義函數(shù)文件costFuncMAD.m，currentBlk、refBlk、 n為傳入?yún)?shù)，cost為返回參數(shù)function cost = costFuncMAD(currentBlk,refBlk, n) % 補充下面程序cost=sum(sum(abs(currentBlk-refBlk)/(n*n); (3)

13、minCost( )% Finds the indices of the cell that holds the minimum cost（找到擁有最小絕對差的點的指數(shù)）% Input% costs : The matrix that contains the estimation costs for a macroblock（包含宏塊的估計代價的矩陣）% Output% dx : the motion vector component in columns（列方向上運動矢量組成）% dy : the motion vector component in rows（行方向上運動矢量組成）% W

14、ritten by Aroh Barjatyafunction dx, dy, min = minCost(costs)row, col = size(costs);% we check whether the current value of costs is less then the already present value in min.% If its inded smaller then we swap the min value with the current one and note the indices.% （檢測costs的當前值是否比已經(jīng)出現(xiàn)的最小值小。如果小的話，

15、我們將當前值與最小值對調(diào)，并注明指數(shù)）% 補充下面程序minnum=65536;x=8;y=8;for i=1:rowfor j=1:colif(costs(i,j)<minnum)minnum=costs(i,j);x=i;y=j;endendenddx=x;dy=y;min=minnum; (4) imgPSNR( )% Computes motion compensated image's PSNR（計算運動補償圖像的峰值信噪比）% Input% imgP : The original image （原始圖像）% imgComp : The compensated image（補償圖像）% n : the peak value possible of any pixel in the images（圖像中任何一個像素的可能的峰值）% Ouput% psnr : The motion compensated image's PSNR（運動補償圖像的峰值信噪比）% Written by Aroh Barjatyafunction psnr = imgPSNR(imgP, imgComp, n)% 補充下面程序MSE=(1/(n*n)*sum(sum(imgP-imgComp).2);PSNR=10*log10(2552/MSE);psnr=PSNR;