matlab对图片进行放大和缩小

利用matlab 对数字图片进行放大缩小是matlab在数字图像处理上的一个简单的应用

matlab库函数imresize()的功能就是这个，那么imresize具体怎么实现的呢，我们可以自己写一个myimresize()

imresize()的用法请查看matlab的HELP，搜索“imresize”

首先我们必须知道彩色数字图像其实是一个m*n*3的数字矩阵组成的，其中的m*n表示图片在宽度和高度上的像素大小，我们通常说的320*240的普通MP3的图片格式就是由宽度上320个像素点和高度上240个像素点组成。而之所以乘3，是因为彩色采用的rgb(red,green,blue)的方式。对于图片上的每个点的颜色，都由3个数字(r,g,b)来决定，如(255,0,0)为红色，（0,255,0）为绿色。每个数字介于0~255之间(8位表示法)。

对图片的放大和缩小，也就是说根据原来的图片矩阵来产生新矩阵。对于新矩阵的每个像素点，其取值有两种方式：

一种是取对应原来位置最近的那个点的像素，这种做法失真比较高，我们称这种做法为“nearest”；

另一种做法是取原来位置周围四个点的加权平均值，这种做法得到的图片比较柔和，我们称这种做法为双线性，“bilinear”，根据数学推导，很容易得知：

f(x0,y0) = (1-b)*[a*f(x+1,y)+(1-a)*f(x,y)] + b*[a*f(x+1,y+1)+(1-a)*f(x,y+1)]

源代码：

function resized = myimresize(image,scale,method);

% A funciton to resize a image

% 'resized' is the result of the function, which means change the size of

% 'image' of 'scale' times with method 'method'

%'image' is the source image

% 'scale' if 'scale'>1 it means to amplify the image with 'scale' times

%       if 'scale' <1 it means to shrink the image with 'scale' times

% 'method' if method = 'nearest' it will find the nearest point of 'image' to

% write into resized

%    else it means the 'bilinear' way

%    All right reserved by Pengxc

if strcmp(method, 'nearest')==1;

    % the first method

    [length,height,layer] = size(image);  % get the basic size of the image

    new_lenth  = length * scale;          % New lenth 

    new_height = height * scale;         % New height 

    new_lenth = floor(new_lenth);        % make it to int

    new_height = floor(new_height);      % make it to int

    

    %The code below is to find the nearest piont 

    for i = 1:new_lenth;

        for j = 1:new_height;

            

            remain_i = (i/scale) - floor(i/scale);

            % To see which side is nearer at x-label

            if remain_i >= 0.5

                o_i = ceil(i/scale);

            else

                o_i = floor(i/scale);

                % When scale>1 and i =1 ,then o_i = 0,which is wrong,so

                % make it equals 1

                if o_i == 0;

                    o_i =1;

                end

            end

            

            remain_j = (j/scale) - floor(i/scale);

            % To see which side is nearer at y-label

            if remain_j >= 0.5

                o_j = ceil(j/scale);

            else

                o_j = floor(j/scale);

                if o_j == 0;

                % When scale>1 and i =1 ,then o_i = 0,which is wrong,so

                % make it equals 1

                    o_j =1;

                end

            end

            

            for k =1:layer

            resized(i,j,k) = image(o_i,o_j,k);

            end

        end

    end

end           

if strcmp(method, 'bilinear')==1;   %    else it means the 'bilinear' way

    [lenth,height,layer] = size(image);  % the same as  above

    new_lenth = lenth * scale;

    new_height = height * scale;

    

    new_lenth = floor(new_lenth);

    new_height = floor(new_height);

    

    for i= 1 : lenth

        for j = 1 :height

            for k = 1: layer

                temp_image(i,j,k) = image(i,j,k);

                %temp_image has a row ans a column more than image

            end

        end

    end

    

    for i=1:lenth

        for k =1:layer

            temp_image(i,height+1,k) = 0;

            % add a column to keep from getting out of matrix 

        end

    end

    

    for j =1:height

        for k=1:layer

            temp_image(lenth+1,j,k)=0;

            % add a row to keep from getting out of matrix 

        end

    end

    

    % The code below use Bilinear to Calulate the value of the resized

    for i=1:new_lenth

        for j =1:new_height

            a = 0;

            b = 0;

            o_i = floor(i/scale);

            o_j = floor(j/scale);

            

            a = (i/scale) - floor(i/scale);

            b = (j/scale) - floor(j/scale);

            %a,b is the parameter, which will be detailly written in the Document

            if o_i == 0;

                o_i = 1;a=0;

            end

            if o_j == 0;

                o_j =1;b=0;

            end

            

            for k =1:layer

                resized(i,j,k) = (1-a)*(1-b)*temp_image(o_i,o_j,k) +(1-a)*b*temp_image(o_i,o_j+1,k) + a*(1-b)*temp_image(o_i+1,o_j,k) +a*b*temp_image(o_i+1,o_j+1,k);

            end

        end

    end    

end

测试与结果：

命令行敲入

I = imread(‘football.jpg’);

MSH=myimresize(I,0.7,’nearest’);

SH = imredize(I,0.7,’nearest’);

imshow(MSH);

figure,imshow(SH);

即可得到结果

左图是本段代码对“nearest”方式缩小0.7倍的测试，右边是库函数