img = imread('face.bmp');       %Assume the 'face.bmp' is in the current directory

%You can even do this.
img = imread('http://www.cse.ucsd.edu/classes/sp04/cse152/hw0/face.bmp'); 
img2 = rgb2gray(img);            %Convert to grayscale
img2 = 255 - double(img2);       %You can add/subtract/multiply/divide a matrix with scalar.
                                 %Note that you need to convert the matrix
                                 %to double. Try remove the conversion and
                                 %see what happen.
img2 = imresize(img2,0.5,'bicubic'); %Resize the image using bicubic interpolation.
imwrite(uint8(img2),'face_small.bmp'); %Save the image. You need to convert it back to uint8. The reason can be seen in the function help:

%help imwrite;
%
%...
%
%Data types
%    ----------
%    Most of the supported image file formats store uint8
%    data.  PNG and TIFF additionally support uint16 data.  For
%    grayscale and RGB images, IF THE DATA ARRAY IS DOUBLE, THE
%    ASSUMED DYNAMIC RANGE IS [0,1].  THE DATA ARRAY IS
%    AUTOMATICALLY SCALED BY 255 BEFORE BEING WRITTEN OUT AS
%    uint8.  If the data array is uint8 or uint16, then it is 
%    written out without scaling as uint8 or uint16, 
%    respectively.  NOTE: If logical data is written to a BMP,
%    PNG or TIFF file, it is assumed to be a binary image and
%    will be written with a bitdepth of 1.
% 
%    For indexed images, if the index array is double, then the
%    indices are first converted to zero-based indices by
%    subtracting 1 from each element, and then they are written
%    out as uint8.  If the index array is uint8 or uint16, 
%    then it is written out without modification as uint8 or 
%    uint16, respectively.
% 
%   .... 

%Another way to do. 
img2 = rgb2gray(img);
img2 = im2double(img2);  %The function 'im2double' is used here 
                         %instead of 'double'. Note the different :

    %IM2DOUBLE takes an image as input, and returns an image of
    %class double.  If the input image is of class double, the
    %output image is identical to it.  If the input image is of
    %class logical, uint8 or uint16, im2double returns the 
    %equivalent image of class double, rescaling or offsetting
    %the data as necessary.

    %DOUBLE Convert to double precision.
    %DOUBLE(X) returns the double precision value for X.
    %If X is already a double precision array, DOUBLE has
    %no effect.

img2 = 1 - img2;
img2 = imresize(img2,0.5,'bicubic');
imwrite(img2,'face_small.bmp')

%Now, just for fun, try this one line solution
imwrite(imresize(1-im2double(rgb2gray(imread('http://www.cse.ucsd.edu/classes/sp04/cse152/hw0/face.bmp'))),0.5,'bic'),'face_small.bmp');