imadjust函数分析一

声明：本文涉及到的行数皆指本文提供的附件imadjust.m的代码中行数

本文只讨论imadjust函数是一种用法，即

J = IMADJUST(I,[LOW_IN; HIGH_IN],[LOW_OUT; HIGH_OUT],GAMMA)

处理效果如下图

图像矩阵I要求数据类型uint8、uint16、double、single和int16，[LOW_IN HIGH_IN]和[LOW_OUT HIGH_OUT] 要求范围都要在[0 1]区间中，不然会出问题。

1.函数首先获得输入参数（行97）

[img,imageType,lowIn,highIn,lowOut,highOut,gamma] = ...
    parseInputs(varargin{:});

根据用法

J = IMADJUST(I,[LOW_IN; HIGH_IN],[LOW_OUT; HIGH_OUT],GAMMA)

其中I赋值给img，imgeType后续有说明，LOW_IN赋值给lowIn，HIGH_IN赋值给highIn，LOW_OUT赋值给lowOUt，HIGH_OUT赋值给highOut，GAMMA赋值给gamma。

进入子函数paraseInputs（行196）

函数会判断输出参数个数多少，本文主要讨论参数个数>1，从行231看起

if nargin == 1    %如果输入参数个数为1
    % IMADJUST(I)
    if ndims(img) ~= 2        % 图像矩阵维数不为2
        eid = sprintf('Images:%s:oneArgOnlyGrayscale',mfilename);
        error(eid, ...
            'IMADJUST(I) is only supported for 2-D grayscale images.');
    end
    % 检验图像矩阵数据类型是否double、uint8、uint16、int16和single
    iptcheckinput(img, {'double' 'uint8' 'uint16' 'int16' 'single'}, ...
        {'2d'}, mfilename, 'I', 1);

    % If a user passes in a m-by-3 double array, assume it is an intensity
    % image (there is really no way to tell).
    imageType = 'intensity';

    % Turn off warning 'Images:imhistc:inputHasNans' before calling STRETCHLIM and
    % restore afterwards. STRETCHLIM calls IMHIST/IMHISTC and the warning confuses
    % a user who calls IMADJUST with NaNs.
    s = warning('off','Images:imhistc:inputHasNaNs');
    lowhigh_in = stretchlim(img);
    warning(s) 

else    % 如果输入参数个数不为1
    if nargin == 2        %如果输入参数个数为2
    % IMADJUST(I,[LOW_IN HIGH_IN])
    % IMADJUST(MAP,[LOW_IN HIGH_IN])
    % IMADJUST(RGB,[LOW_IN HIGH_IN])
    if ~isempty(varargin{2})    %如果输入参数2的数组不为空
        lowhigh_in = varargin{2};
    end

    elseif nargin == 3    %如果输入参数个数为3
        % IMADJUST(I,[LOW_IN HIGH_IN],[LOW_OUT HIGH_OUT])
        % IMADJUST(MAP,[LOW_IN HIGH_IN],[LOW_OUT HIGH_OUT])
        % IMADJUST(RGB,[LOW_IN HIGH_IN],[LOW_OUT HIGH_OUT])

        if ~isempty(varargin{2})
            lowhigh_in = varargin{2};
        end
        if ~isempty(varargin{3})
            lowhigh_out = varargin{3};
        end
    else    %如果输入参数个数为4
        % IMADJUST(I,[LOW_IN HIGH_IN],[LOW_OUT HIGH_OUT],GAMMA)
        % IMADJUST(MAP,[LOW_IN HIGH_IN],[LOW_OUT HIGH_OUT],GAMMA)
        % IMADJUST(RGB,[LOW_IN HIGH_IN],[LOW_OUT HIGH_OUT],GAMMA)
        if ~isempty(varargin{2})
            lowhigh_in = varargin{2};
        end
        if ~isempty(varargin{3})
            lowhigh_out = varargin{3};
        end
        if ~isempty(varargin{4})
            gamma = varargin{4};
        end
    end
    imageType = findImageType(img, lowhigh_in, lowhigh_out);
    checkRange(lowhigh_in, imageType, 2,'[LOW_IN; HIGH_IN]');
    checkRange(lowhigh_out, imageType, 3,'[LOW_OUT; HIGH_OUT]');
end

这个if-else语句代码很明显看到函数输入参数[LOW_IN HIGH_IN]赋值给了lowhigh_in，参数[LOW_OUT HIGH_OUT]赋值给了lowhigh_out，而参数GAMMA赋值给了gamma，然后执行参数有效性判断。（行265）

imageType = findImageType(img, lowhigh_in, lowhigh_out);
checkRange(lowhigh_in, imageType, 2,'[LOW_IN; HIGH_IN]');
checkRange(lowhigh_out, imageType, 3,'[LOW_OUT; HIGH_OUT]');

第一个语句主要用于根据输入图像矩阵判断图像类型，第二个和第三个主要是判断[LOW_IN HIGH_IN]和[LOW_OUT HIGH_OUT]的正确性。

（1）findImageType函数如下（行274）

function imageType = findImageType(img, lowhigh_in, lowhigh_out)

if (ndims(img)==3 && size(img,3)==3)
    % RGB image
    % RGB图像 第三维的维数为3
    iptcheckinput(img, {'double' 'uint8' 'uint16' 'int16' 'single'}, ...
        {}, mfilename, 'RGB1', 1);
    imageType = 'truecolor';

elseif (numel(lowhigh_in) == 2 && numel(lowhigh_out) == 2) || ...
    size(img,2) ~= 3
    % Assuming that a user passed in an intensity image if lowhigh_in and
    % lowhigh_out are 2-element vectors, e.g., imadjust(3 column image,
    % [1;2], [2;3]).
    % 输入图像矩阵维数中列数不为3
    iptcheckinput(img, {'double' 'uint8' 'uint16' 'int16' 'single'}, ...
        {'2d'}, mfilename, 'I', 1);
    imageType = 'intensity';

else
    %Colormap
    iptcheckmap(img,mfilename,'MAP',1);
    imageType = 'indexed';
end

图像类型分为三种，第一种是'truecolor'图像，第二种是'intensity'，第三种为'indexed'。本文讨论的为第二种（一般情况使用的就是第二种，其他的目前没遇到，后续文章更新），即'intensity'，判定第二种条件为[LOW_IN HIGH_IN]和[LOW_OUT HIGH_OUT]的元素个数为2并且输入图像矩阵列数不为3。

（2）chackRange函数如下（行300）

function checkRange(range, imageType, argumentPosition, variableName)

if strcmp(imageType, 'intensity')
    if numel(range) ~= 2
        eid = sprintf('Images:%s:InputMustBe2ElVec',mfilename);
        error(eid, ...
            'Function %s expected its %s input argument, %s\n%s', ...
            mfilename, iptnum2ordinal(argumentPosition), variableName, ...
            'to be a two-element vector.');
    end
else
    if ~(numel(range) == 2 || isequal(size(range), [2 3]))
        eid = sprintf('Images:%s:InputMustBe2ElVecOr2by3Matrix',mfilename);
        error(eid, ...
            'Function %s expected its %s input argument, %s\n%s', ...
            mfilename, iptnum2ordinal(argumentPosition), variableName, ...
            'to be a two-element vector or a 2-by-3 matrix.');
    end
end

前面判定图像为'intensity'，这里主要判断[LOW_IN HIGH_IN]和[LOW_OUT HIGH_OUT]元素个数必须为2，不然报错，这里就是第一个检查，后面还有检查。

最后

[low_in high_in]   = splitRange(lowhigh_in, imageType);
[low_out high_out] = splitRange(lowhigh_out, imageType);

获取参数[low_in high_in] 和 [low_out high_out] ，然后函数返回。

2.参数检查

前面通过函数paraseInputs获取了用户输入的参数，接下来进行参数有效性检验

validateLowHigh(lowIn,highIn,lowOut,highOut);
gamma = validateGamma(gamma,imageType);

（1）函数validateLowHigh（行340）

function validateLowHigh(lowIn,highIn,lowOut,highOut)

if any(lowIn >= highIn)
    eid = sprintf('Images:%s:lowMustBeSmallerThanHigh',mfilename);
    error(eid, '%s: LOW_IN must be less than HIGH_IN.',...
        upper(mfilename));
end

if isInvalidRange(lowIn) || isInvalidRange(highIn) ...
        || isInvalidRange(lowOut) || isInvalidRange(highOut)
    eid = sprintf('Images:%s:parametersAreOutOfRange',mfilename);
    error(eid, '%s: LOW_IN, HIGH_IN, LOW_OUT and HIGH_OUT %s',...
        upper(mfilename), 'must be in the range [0.0, 1.0].');
end

该函数首先检查了lowIn要小于highIn，但是没要求lowOut要小于highOut，因为这个函数有个运用，在于可以反转图像，即[0 1]变换到[1 0]。

然后使用函数isInvalidRange限制了输入lowIn、highIn、lowOut和highOut范围为[0 1]，这就是开头所说[LOW_IN HIGH_IN]和[LOW_OUT HIGH_OUT] 要求范围都要在[0 1]区间中的出处。

（2）函数validateGamma（行362）

function gamma = validateGamma(gamma,image_type)

if strcmp(image_type,'intensity')
    iptcheckinput(gamma,{'double'},{'scalar', 'nonnegative'}, ...
        mfilename, 'GAMMA', 4)
else
    iptcheckinput(gamma,{'double'},{'nonnegative','2d'},...
        mfilename, 'GAMMA', 4)
    if numel(gamma) == 1,
        gamma = gamma*ones(1,3);
    end
end

前面判定图像类型为'intensity'，这里判定gamma参数必须为正数。

3.执行变换（行104）

if ~isfloat(img) && numel(img) > 65536
    % integer data type image with more than 65536 elements
    out = adjustWithLUT(img,lowIn,highIn,lowOut,highOut,gamma);

else
    classin = class(img);
    classChanged = false;
    if ~isa(img,'double')    %如果图像矩阵数据类型不为double, 图像矩阵数据范围转为[0,1]
        classChanged = true;
        img = im2double(img);
    end

    if strcmp(imageType, 'intensity')
        out = adjustGrayscaleImage(img,lowIn,highIn,lowOut,highOut,gamma);
    elseif strcmp(imageType, 'indexed')
        out = adjustColormap(img,lowIn,highIn,lowOut,highOut,gamma);
    else
        out = adjustTruecolorImage(img,lowIn,highIn,lowOut,highOut,gamma);
    end

    if classChanged
        out = changeclass(classin,out);
    end

end

首先判断如果图像矩阵数据不为浮点数（double和single）并且矩阵元素个数>65536，则执行函数adjustWithLUT函数。

（1）adjustWithLUT函数如下

function out = adjustWithLUT(img,lowIn,highIn,lowOut,highOut,gamma)

imgClass = class(img);
out = zeros(size(img),imgClass);

%initialize for lut

switch imgClass
    case 'uint8'
        lutLength = 256;
        conversionFcn = @im2uint8;
    case 'uint16'
        lutLength = 65536;
        conversionFcn = @im2uint16;
    case 'int16'
        lutLength = 65536;
        conversionFcn = @im2int16;
    otherwise
        eid = sprintf('Images:%s:internalError',mfilename);
        msg = 'Internal error: invalid class type.';
        error(eid,'%s',msg);
end

关键部分

for p = 1:size(img,3)
    lut = linspace(0,1,lutLength);
    scalingFactor = 1;
    lut = adjustArray(lut,lowIn(p),highIn(p),lowOut(p),highOut(p), ...
        gamma(p),scalingFactor);
    lut = conversionFcn(lut);
    out(:,:,p) = intlut(img(:,:,p),lut);
end

首先创建一个数组作为筛选数组，不需要对图像矩阵进行变换，直接根据筛选数组的筛选图像矩阵。以下语句为灰度值归一化，将灰度归一化到[0 1]范围，这是为了后面变换方便。

lut = linspace(0,1,lutLength);

lutLength的大小为图像矩阵数据类型对应该类型的最大值，例如图像矩阵数据类型为uint8，则lutLength=256。该语句作用在于创建一个数组lut，数组储存归一化后的灰度值，每个元素分别对于相应的灰度等级。

lut = adjustArray(lut,lowIn(p),highIn(p),lowOut(p),highOut(p), ...
        gamma(p),scalingFactor);

对lut进行变换，变换结果如开头的图

lowIn=低输入 highIn=高输入 lowOut=低输出 highOut=高输出

由于数组lut元素范围[0 1]，这是归一化数据，因此要还原对应的值

lut = conversionFcn(lut);

剩下就是筛选了

out(:,:,p) = intlut(img(:,:,p),lut);

（2）adjustArray函数（行187）

function out = adjustArray(img,lIn,hIn,lOut,hOut,g,d)

%make sure img is in the range [lIn;hIn]
img(:) =  max(lIn(d,:), min(hIn(d,:),img));

out = ( (img - lIn(d,:)) ./ (hIn(d,:) - lIn(d,:)) ) .^ (g(d,:));
out(:) = out .* (hOut(d,:) - lOut(d,:)) + lOut(d,:);

首先筛选掉矩阵img中小于lIn的数的数值变为lIn，大于hIn的数的数值变为hIn，筛选语句如下

img(:) =  max(lIn(d,:), min(hIn(d,:),img));

接着进行变换，令 并且d=1，则该函数处理的效果等价于：

至于为什么这样，我手写推导过程如下

表达式就是曲线AB的方程

附件：imadjust.m代码

 function out = imadjust(varargin)
 %IMADJUST Adjust image intensity values or colormap.
 %   J = IMADJUST(I) maps the values in intensity image I to new values in J
 %   such that 1% of data is saturated at low and high intensities of I.
 %   This increases the contrast of the output image J.
 %
 %   J = IMADJUST(I,[LOW_IN; HIGH_IN],[LOW_OUT; HIGH_OUT]) maps the values
 %   in intensity image I to new values in J such that values between LOW_IN
 %   and HIGH_IN map to values between LOW_OUT and HIGH_OUT. Values below
 %   LOW_IN and above HIGH_IN are clipped; that is, values below LOW_IN map
 %   to LOW_OUT, and those above HIGH_IN map to HIGH_OUT. You can use an
 %   empty matrix ([]) for [LOW_IN; HIGH_IN] or for [LOW_OUT; HIGH_OUT] to
 %   specify the default of [0 1]. If you omit the argument, [LOW_OUT;
 %   HIGH_OUT] defaults to [0 1].
 %
 %   J = IMADJUST(I,[LOW_IN; HIGH_IN],[LOW_OUT; HIGH_OUT],GAMMA) maps the
 %   values of I to new values in J as described in the previous syntax.
 %   GAMMA specifies the shape of the curve describing the relationship
 %   between the values in I and J. If GAMMA is less than 1, the mapping is
 %   weighted toward higher (brighter) output values. If GAMMA is greater
 %   than 1, the mapping is weighted toward lower (darker) output values. If
 %   you omit the argument, GAMMA defaults to 1 (linear mapping).
 %
 %   NEWMAP = IMADJUST(MAP,[LOW_IN; HIGH_IN],[LOW_OUT; HIGH_OUT],GAMMA)
 %   transforms the colormap associated with an indexed image. If LOW_IN,
 %   HIGH_IN, LOW_OUT, HIGH_OUT, and GAMMA are scalars, then the same
 %   mapping applies to red, green and blue components. Unique mappings for
 %   each color component are possible when: LOW_IN and HIGH_IN are both
 %   1-by-3 vectors, LOW_OUT and HIGH_OUT are both 1-by-3 vectors, OR GAMMA
 %   is a 1-by-3 vector. The rescaled colormap, NEWMAP, is the same size as
 %   MAP.
 %
 %   RGB2 = IMADJUST(RGB1,...) performs the adjustment on each image plane
 %   (red, green, and blue) of the RGB image RGB1. As with the colormap
 %   adjustment, you can apply unique mappings to each plane.
 %
 %   Note that IMADJUST(I) is equivalent to IMADJUST(I,STRETCHLIM(I)).
 %
 %   Note that if HIGH_OUT < LOW_OUT, the output image is reversed, as in a
 %   photographic negative.
 %
 %   Class Support
 %   -------------
 %   For syntaxes that include an input image (rather than a colormap), the
 %   input image can be uint8, uint16, int16, double, or single. The output
 %   image has the same class as the input image. For syntaxes that include
 %   a colormap, the input and output colormaps are double.
 %
 %   Examples
 %   --------
 %       I = imread('pout.tif');
 %       J = imadjust(I);
 %       figure, imshow(I), figure, imshow(J)
 %
 %       K = imadjust(I,[0.3 0.7],[]);
 %       figure, imshow(K)
 %
 %       RGB1 = imread('football.jpg');
 %       RGB2 = imadjust(RGB1,[.2 .3 0; .6 .7 1],[]);
 %       figure, imshow(RGB1), figure, imshow(RGB2)
 %
 %   See also BRIGHTEN, DECORRSTRETCH, HISTEQ, IMCONTRAST, STRETCHLIM.

 %   Copyright 1992-2007 The MathWorks, Inc.
 %   $Revision: 5.26.4.9 $  $Date: 2007/12/10 21:37:18 $

 %   Input-output specs
 %   ------------------
 %   I,J          real, full matrix, 2-D
 %                uint8, uint16, double, single, int16
 %
 %   RGB1,RGB2    real, full matrix
 %                M-by-N-by-3
 %                uint8, uint16, double, single, int16
 %
 %   MAP,NEWMAP   real, full matrix
 %                M-by-3
 %                double with values in the range [0,1].
 %
 %   [LOW_IN; HIGH_IN]    double, real, full matrix
 %                        For I, size can only be 2 elements.
 %                        For RGB or MAP, size can be 2 elements OR
 %                        2-by-3 matrix.
 %                        LOW_IN < HIGH_IN
 %
 %   [LOW_OUT; HIGH_OUT]  Same size restrictions as [LOW_IN; HIGH_IN]
 %                        LOW_OUT can be less than HIGH_OUT
 %
 %   LOW_IN, HIGH_IN, LOW_OUT, HIGH_OUT all must be in the range [0,1];
 %
 %   GAMMA         real, double, nonnegative
 %                 scalar for I
 %                 scalar or 1-by-3 vector for RGB and MAP

 %Parse inputs and initialize variables
 [img,imageType,lowIn,highIn,lowOut,highOut,gamma] = ...
     parseInputs(varargin{:});

 validateLowHigh(lowIn,highIn,lowOut,highOut);
 gamma = validateGamma(gamma,imageType);

 if ~isfloat(img) && numel(img) > 65536
     % integer data type image with more than 65536 elements
     out = adjustWithLUT(img,lowIn,highIn,lowOut,highOut,gamma);

 else
     classin = class(img);
     classChanged = false;
     if ~isa(img,'double')    %如果图像矩阵数据类型不为double, 图像矩阵数据范围转为[0,1]
         classChanged = true;
         img = im2double(img);
     end

     if strcmp(imageType, 'intensity')
         out = adjustGrayscaleImage(img,lowIn,highIn,lowOut,highOut,gamma);
     elseif strcmp(imageType, 'indexed')
         out = adjustColormap(img,lowIn,highIn,lowOut,highOut,gamma);
     else
         out = adjustTruecolorImage(img,lowIn,highIn,lowOut,highOut,gamma);
     end

     if classChanged
         out = changeclass(classin,out);
     end

 end

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 function out = adjustWithLUT(img,lowIn,highIn,lowOut,highOut,gamma)

 imgClass = class(img);
 out = zeros(size(img),imgClass);

 %initialize for lut

 switch imgClass
     case 'uint8'
         lutLength = 256;
         conversionFcn = @im2uint8;
     case 'uint16'
         lutLength = 65536;
         conversionFcn = @im2uint16;
     case 'int16'
         lutLength = 65536;
         conversionFcn = @im2int16;
     otherwise
         eid = sprintf('Images:%s:internalError',mfilename);
         msg = 'Internal error: invalid class type.';
         error(eid,'%s',msg);
 end

 for p = 1:size(img,3)
     lut = linspace(0,1,lutLength);
     scalingFactor = 1;
     lut = adjustArray(lut,lowIn(p),highIn(p),lowOut(p),highOut(p), ...
         gamma(p),scalingFactor);
     lut = conversionFcn(lut);
     out(:,:,p) = intlut(img(:,:,p),lut);
 end

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 function out = adjustColormap(cmap,lIn,hIn,lOut,hOut,g)

 % expansion factor that can expand a 1-by-3 range to the size of cmap.
 expansionFactor = ones(size(cmap,1), 1);
 out = adjustArray(cmap, lIn, hIn, lOut, hOut, g, expansionFactor);

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 function out = adjustGrayscaleImage(img,lIn,hIn,lOut,hOut,g)

 expansionFactor = 1;
 out = adjustArray(img, lIn, hIn, lOut, hOut, g, expansionFactor);

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 function out = adjustTruecolorImage(rgb,lIn,hIn,lOut,hOut,g)

 out = zeros(size(rgb), class(rgb));
 expansionFactor = 1;
 for p = 1 : 3
     out(:,:,p) = adjustArray(rgb(:,:,p), lIn(p),hIn(p), lOut(p), ...
         hOut(p), g(p), expansionFactor);
 end

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 function out = adjustArray(img,lIn,hIn,lOut,hOut,g,d)

 %make sure img is in the range [lIn;hIn]
 img(:) =  max(lIn(d,:), min(hIn(d,:),img));

 out = ( (img - lIn(d,:)) ./ (hIn(d,:) - lIn(d,:)) ) .^ (g(d,:));
 out(:) = out .* (hOut(d,:) - lOut(d,:)) + lOut(d,:);

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 % 参数获取
 function [img,imageType,low_in,high_in,low_out,high_out,gamma] = ...
     parseInputs(varargin)

 iptchecknargin(1,4,nargin,mfilename);    % 输入参数个数在1-4个
 img = varargin{1};    % img保存图像矩阵

 % Default values (默认值)
 lowhigh_in  = [0; 1];
 lowhigh_out = [0; 1];
 gamma = 1;

 if nargin == 1    %如果输入参数个数为1
     % IMADJUST(I)
     if ndims(img) ~= 2        % 图像矩阵维数不为2
         eid = sprintf('Images:%s:oneArgOnlyGrayscale',mfilename);
         error(eid, ...
             'IMADJUST(I) is only supported for 2-D grayscale images.');
     end
     % 检验图像矩阵数据类型是否double、uint8、uint16、int16和single
     iptcheckinput(img, {'double' 'uint8' 'uint16' 'int16' 'single'}, ...
         {'2d'}, mfilename, 'I', 1);

     % If a user passes in a m-by-3 double array, assume it is an intensity
     % image (there is really no way to tell).
     imageType = 'intensity';

     % Turn off warning 'Images:imhistc:inputHasNans' before calling STRETCHLIM and
     % restore afterwards. STRETCHLIM calls IMHIST/IMHISTC and the warning confuses
     % a user who calls IMADJUST with NaNs.
     s = warning('off','Images:imhistc:inputHasNaNs');
     lowhigh_in = stretchlim(img);
     warning(s) 

 else    % 如果输入参数个数不为1
     if nargin == 2        %如果输入参数个数为2
     % IMADJUST(I,[LOW_IN HIGH_IN])
     % IMADJUST(MAP,[LOW_IN HIGH_IN])
     % IMADJUST(RGB,[LOW_IN HIGH_IN])
     if ~isempty(varargin{2})    %如果输入参数2的数组不为空
         lowhigh_in = varargin{2};
     end

     elseif nargin == 3    %如果输入参数个数为3
         % IMADJUST(I,[LOW_IN HIGH_IN],[LOW_OUT HIGH_OUT])
         % IMADJUST(MAP,[LOW_IN HIGH_IN],[LOW_OUT HIGH_OUT])
         % IMADJUST(RGB,[LOW_IN HIGH_IN],[LOW_OUT HIGH_OUT])

         if ~isempty(varargin{2})
             lowhigh_in = varargin{2};
         end
         if ~isempty(varargin{3})
             lowhigh_out = varargin{3};
         end
     else    %如果输入参数个数为4
         % IMADJUST(I,[LOW_IN HIGH_IN],[LOW_OUT HIGH_OUT],GAMMA)
         % IMADJUST(MAP,[LOW_IN HIGH_IN],[LOW_OUT HIGH_OUT],GAMMA)
         % IMADJUST(RGB,[LOW_IN HIGH_IN],[LOW_OUT HIGH_OUT],GAMMA)
         if ~isempty(varargin{2})
             lowhigh_in = varargin{2};
         end
         if ~isempty(varargin{3})
             lowhigh_out = varargin{3};
         end
         if ~isempty(varargin{4})
             gamma = varargin{4};
         end
     end
     imageType = findImageType(img, lowhigh_in, lowhigh_out);
     checkRange(lowhigh_in, imageType, 2,'[LOW_IN; HIGH_IN]');
     checkRange(lowhigh_out, imageType, 3,'[LOW_OUT; HIGH_OUT]');
 end

 [low_in high_in]   = splitRange(lowhigh_in, imageType);
 [low_out high_out] = splitRange(lowhigh_out, imageType);

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 function imageType = findImageType(img, lowhigh_in, lowhigh_out)

 if (ndims(img)==3 && size(img,3)==3)
     % RGB image
     % RGB图像 第三维的维数为3
     iptcheckinput(img, {'double' 'uint8' 'uint16' 'int16' 'single'}, ...
         {}, mfilename, 'RGB1', 1);
     imageType = 'truecolor';

 elseif (numel(lowhigh_in) == 2 && numel(lowhigh_out) == 2) || ...
     size(img,2) ~= 3
     % Assuming that a user passed in an intensity image if lowhigh_in and
     % lowhigh_out are 2-element vectors, e.g., imadjust(3 column image,
     % [1;2], [2;3]).
     % 输入图像矩阵维数中列数不为3
     iptcheckinput(img, {'double' 'uint8' 'uint16' 'int16' 'single'}, ...
         {'2d'}, mfilename, 'I', 1);
     imageType = 'intensity';

 else
     %Colormap
     iptcheckmap(img,mfilename,'MAP',1);
     imageType = 'indexed';
 end

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 function checkRange(range, imageType, argumentPosition, variableName)

 if strcmp(imageType, 'intensity')
     if numel(range) ~= 2
         eid = sprintf('Images:%s:InputMustBe2ElVec',mfilename);
         error(eid, ...
             'Function %s expected its %s input argument, %s\n%s', ...
             mfilename, iptnum2ordinal(argumentPosition), variableName, ...
             'to be a two-element vector.');
     end
 else
     if ~(numel(range) == 2 || isequal(size(range), [2 3]))
         eid = sprintf('Images:%s:InputMustBe2ElVecOr2by3Matrix',mfilename);
         error(eid, ...
             'Function %s expected its %s input argument, %s\n%s', ...
             mfilename, iptnum2ordinal(argumentPosition), variableName, ...
             'to be a two-element vector or a 2-by-3 matrix.');
     end
 end

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 function [rangeMin rangeMax] = splitRange(range, imageType)

 if numel(range) == 2
     if strcmp(imageType, 'intensity')
         rangeMin = range(1);
         rangeMax = range(2);
     else
         % Create triples for RGB image or Colormap
         rangeMin = range(1) * ones(1,3);
         rangeMax = range(2) * ones(1,3);
     end
 else
     % range is a 2 by 3 array
     rangeMin = range(1,:);
     rangeMax = range(2,:);
 end

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 function validateLowHigh(lowIn,highIn,lowOut,highOut)

 if any(lowIn >= highIn)
     eid = sprintf('Images:%s:lowMustBeSmallerThanHigh',mfilename);
     error(eid, '%s: LOW_IN must be less than HIGH_IN.',...
         upper(mfilename));
 end

 if isInvalidRange(lowIn) || isInvalidRange(highIn) ...
         || isInvalidRange(lowOut) || isInvalidRange(highOut)
     eid = sprintf('Images:%s:parametersAreOutOfRange',mfilename);
     error(eid, '%s: LOW_IN, HIGH_IN, LOW_OUT and HIGH_OUT %s',...
         upper(mfilename), 'must be in the range [0.0, 1.0].');
 end

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 function isInvalid = isInvalidRange(range)

 isInvalid = min(range) < 0 || max(range) > 1;

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 function gamma = validateGamma(gamma,image_type)

 if strcmp(image_type,'intensity')
     iptcheckinput(gamma,{'double'},{'scalar', 'nonnegative'}, ...
         mfilename, 'GAMMA', 4)
 else
     iptcheckinput(gamma,{'double'},{'nonnegative','2d'},...
         mfilename, 'GAMMA', 4)
     if numel(gamma) == 1,
         gamma = gamma*ones(1,3);
     end
 end