Errata

The code samples that referenced in book are missing under the Essential Links section.

In chapter two, there are a number of small errors that affect the code examples, details follow. Please note that that code on GitHub reflects these changes already.

-----

• Example 2-1:
○ Change "Example1" to "Example 2-1" p23
§ Note that this appears in three places in the code. Also, where the code is described in detail (pages 24-25), each of these three lines appear again, and this substitution should be made in each of these locations (once on p24, twice on p25). It also appears in the text at the bottom of p24, this also should be changed.
• Example 2-2:
○ Change "Example1" to "Example 2-2" p24
§ Note that this appears in three places in the code.
• Example 2-3:
○ Change "Example3" to "Example 2-3" p26
§ Also note that, this substitution must also be made at the middle of p26, just after "in this case, named...", and again lower on page 26, this same substitution needs to be made again in the code 'snipet' just above the words "Once inside of…".
○ Insert a line reading "using namespace std;" after the second "#include" line and before the line beginning with "int main(…" p26
○ Change "cv::waitKey(33)" to "(char) cv::waitKey(33)" p26
• Example 2-4:
○ Change "Example2-4" to "Example 2-4" p28
§ Note that this appears in four places in the code.
§ again in code 'snip' in the middle of p30
§ again twice in code 'snip' at top of p30
○ Change "cv::CAP_PROP_FRAME_COUNT" to "CV_CAP_PROP_FRAME_COUNT" p27
§ again in code 'snip' at top of p30
○ Change "cv::CAP_PROP_FRAME_WIDTH" to "CV_CAP_PROP_FRAME_WIDTH" p27
§ again in code 'snip' at top of p30
○ Change "cv::CAP_PROP_FRAME_HEIGHT" to "CV_CAP_PROP_FRAME_HEIGHT" p27
○ Change "cv::CAP_PROP_POS_FRAMES" to "CV_CAP_PROP_POS_FRAMES" p27
§ again in code 'snip' at bottom of p29
§ again in text at top of p30
§ again in code 'snip' at top of p30
• Example 2-5:
○ Change "Example4-out" to "Example 2-5-out" at bottom of p31
○ Change "Example2_5-in" to "Example 2-5-in" p32
§ Note that this appears in two places in the code.
○ Change "Example2_5-out" to "Example 2-5-out" p32
§ Note that this appears in two places in the code.
○ On p32, the line reading "void example2_5( const cv::Mat & image ) {" should be changed to "int main( int argc, char** argv ) {" should be replaced by the following give lines: (note that second line is blank, third through fifth lines are double space indent.)
"""
int main( int argc, char** argv ) {

// Load an image specified on the command line.
//
cv::Mat image = cv::imread(argv[1],-1);
"""
• Example 2-6:
○ "img = cv::imread…" should be "img1 = cv::imread…"
○ Change "Example1" to "Example 2-6-in" p33
§ Note that this appears in two places in the code.
○ Change "Example2" to "Example 2-6-out" p33
§ Note that this appears in two places in the code.
• Example 2-10:
○ Change "Example2_10" to "Example 2-10" p35
• Example 2-11:
○ Change "Example2_11" to "Example 2-11" p36, p37
§ Note that this appears in two places in the code.
○ Change "cv::WARP_FILL_OUTLIERS" to "CV_WARP_FILL_OUTLIERS" p37
○ Change "cv::CAP_PROP_FRAME_FPS" to "CV_CAP_PROP_FRAME_FPS" p36
○ Change "cv::CAP_PROP_FRAME_WIDTH" to "CV_CAP_PROP_FRAME_WIDTH" p36
○ Change "cv::CAP_PROP_FRAME_HEIGHT" to "CV_CAP_PROP_FRAME_HEIGHT" p37
○ Change "cv::waitKey(10)" to "(char) cv::waitKey(33)"

The text "Once inside of the while() loop" should be changed to "Once inside of the main loop". Please note that the word "while()" is in 'inline code' font, while the replacement word 'main' should be in 'normal body' font.

the variable intensity is set, but not used in the example.

from context, I cannot guess how it might be used. (Sorry, Im Learning OpenCV3)


//***New issue in same example:::
on these lines:
cout << "The Gray pixel there is: " << (unsigned int)img_gray.at<uchar>(x,y) << "\n";
x/=4;
y/=4;
cout << "Pyramid pixel there is: " << (unsigned int)img_pyr1.at<uchar>(x,y) << "\n";

x,y are reversed y,x in the book, example2-9
perhaps it was intentional, but Im thinking not.

Note from the Author or Editor:
tl;dr: One small change, error is actually in the comment.

Relating to the "intensity" variable: this is just pedagogical, it was intended to make clear that the intensity, as a vector, can be extracted with using at<cv::Vec3b>, while later on, the use of at<uchar> is shown as an alternative.

at<> access and "y, x": The correct access is image.at<uchar>(y,x). (The logic here has to do with the "running speed" of an index, i.e. if you were to access a pixel in a byte array, you might do something like: pixel = y*row_size+x. And in a three dimensional array z*plane_size+y*row_size+x. So this is in keeping with the idea of a "base-N" number as a way of thinking about an image or image tensor.)

However: There is an error, which you might not have noticed, which is that in the comment at the bottom or page 34, the accessor actually does say "img_rgb.at< cv::Vec3b >(x,y)[0]". It is the comment that is wrong. This should be changed to read "y,x" instead of "x,y".

it is missing the prefix "COLOR_" when converting the image color:

cv::cvtColor( img_rgb, img_gry, cv::BGR2GRAY );

Note from the Author or Editor:
The submitter is correct.

* Please change "cv::BGR2GRAY" to "cv::COLOR_BGR2GRAY" in the middle of p34.

Point2f belongs to the namespace cv, and as such it should be referred to as cv::Point2f. The code as it is will not compile since Point2f is undeclared in the scope of the main function.

In other words, the 12th line of code of page 37 should read

cv::Point2f( // Centerpoint for log-polar transformation

instead of

Point2f( // Centerpoint for log-polar transformation

Sentence suggested the existence of types Point2s and Point3s based on 'short' primitive, which there are not (at least in OpenCV 3.2)

Note from the Author or Editor:
Thank you. This is correct. The text does make it seem like there should be such a type.

* Delete text ", s is a short integer" from comma separated list.

Conversely to what is there stated, there's no std::smart_ptr<> class in C++ STL. Indeed, cv::Ptr<> has similarities with std::shared_ptr<> class (introduced in C++11 standard).

Note from the Author or Editor:
The submitter is correct. The text should read "std::shared_ptr<>", not "std::smart_ptr<>" (as it does in the case for the Boost library).

* Change "std::shared_ptr<>" to "std::smart_ptr<>" in footnote 7.

in OpenCV3, cxtypes.h doesn't exist. ( I recognize this was a version 1 header.)
the modern replacement, would be "opencv2/core/types_c.h"

If this varies, on windows machines, I have no way to check, but in Linux:

$ locate cxtypes.h
$ locate types_c.h
~/Downloads/opencv-3.2.0/modules/core/include/opencv2/core/types_c.h
~/Downloads/opencv-3.2.0/modules/imgproc/include/opencv2/imgproc/types_c.h
~/opencv/build/junk/types_c.h.junk
~/opencv/modules/core/include/opencv2/core/types_c.h
~/opencv/modules/imgproc/include/opencv2/imgproc/types_c.h
/usr/include/opencv2/core/types_c.h
/usr/include/opencv2/imgproc/types_c.h
/usr/local/include/opencv2/core/types_c.h
/usr/local/include/opencv2/imgproc/types_c.h

Note from the Author or Editor:
The reader is exactly correct. the text ".../opencv/cxcore/include/cxtypes.h" should be changed to read "...opencv/modules/core/include/opencv2/core/types_c.h"

The comment for the 3rd parameter (dim) of the cv::reduce() prototype says: "Reduction direction (1) to row, (0) to col"
However, the table below, and the official documentation says the opposite.

The following definition, p145 on Dec2016 verson,
cv::repeat()
void cv::repeat(
cv::InputArray src, // Input 2-dimensional array
int nx, // Copies in x-direction
int ny, // Copies in y-direction
cv::OutputArray dst // Result array
);
cv::Mat cv::repeat( // Return result array
cv::InputArray src, // Input 2-dimensional array
int nx, // Copies in x-direction
int ny // Copies in y-direction
);

I was using this function to build an image mask and found that the width and height values of the resultant matrix were transposed. OpenCV web page shows that the function parameters nx and ny should be the other way round ny, nx as follows.

Mat repeat(const Mat& src, int ny, int nx) (C++ function, in Operations on Arrays)
void repeat(InputArray src, int ny, int nx, OutputArray dst) (C++ function, in Operations on Arrays)

Note from the Author or Editor:
This is correct. The lines for Nx and Ny are incorrectly transposed in the text.

* Please reverse the order of the lines containing "int nx, // Copies in x-direction" and "int ny, // Copies in y-direction" in the code describing cv::repeat() at the bottom of the page. Please note that the error appears twice, once in each prototype for cv::repeat(). p145

FileStorage::FileStorage();
should be
cv：：FileStorage::FileStorage();

It's more clear.

Note from the Author or Editor:
Reader is correct

* Please change "FileStorage::FileStorage" to "cv::FileStorage::FileStorage" in code example 'snips'. Please note that this correction needs to be made three times, once right after "Writing to a cv::FileStorage", and twice just before "Once you have opened...". p198

The book says cv::namedWindow() takes cv::GUI_NORMAL, cv::GUI_EXTENDED as second parameter.
But in reality, those are cv::WINDOW_GUI_NORMAL and cv::WINDOW_GUI_EXPANDED.

Note from the Author or Editor:
Comment is only partially correct. The correct usage is to use CV_WINDOW_NORMAL or CV_WINDOW_AUTOSIZE to determine how the window is sized, and CV_GUI_NORMAL and CV_GUI_AUTOSIZE to determine if the enhanced Qt based interface is to be enabled. It should be noted that in the case of the former two, it is also acceptable (and even preferred) to use cv::WINDOW_NORMAL or cv::WINDOW_AUTOSIZE (i.e. the cv:: namespace versions), but for the latter pair, the library has not yet been modified so as to make in "in namespace" version available.

However, there is a different error, which is that the book says cv::GUI_EXTENDED when the correct usage is CV_GUI_EXPANDED. Note that The book says "EXTENDED" when the correct flag is "EXPANDED", but there is also the more subtle issue of the "cv::" namespace version, described in the book, not yet being released in the library.

* Change cv::GUI_EXTENDED to CV_GUI_EXPANDED in both places it appears on p222, and in the third paragraph on p224. Also change cv::GUI_NORMAL to CV_GUI_NORMAL in the second paragraph on p222, and in the first paragraph on p223.

To register OpenGL callback function,
the book teaches to use cv:createOpenGLCallback() function.
But, in the OpenCV reference site on OPpenCV 3.2.0,
there is no such function.
In the OpenCV homepage, there is just setOpenGlDrawCallback() function.

Note from the Author or Editor:
User is correct.

* Change "createOpenGLCallback() to setOpenGLCallback()" where it appears in the second paragraph and in the second code 'snip' on p231.

To build example 10-3, you have to add the following lines, or it won't compile:
At top (missing):
#include <opencv2/opencv.hpp>
#include <cstdlib>

And a constant has changed so that you must replace:
cv::Mat Igray = cv::imread(argv[6], cv::LOAD_IMAGE_GRAYSCALE);
with:
cv::Mat Igray = cv::imread(argv[6], cv::IMREAD_GRAYSCALE);

cv::calcHist syntax is repeated twice
second C and also STL C++ are missing

Note from the Author or Editor:
The submitter is incorrect that the same text is repeated twice, one prototype applies to the case in which the output is a dense array ("OutputArray"), while the other applies to the case in which the output is a sparse array ("SpraseMat").

However, the submitter *is correct* that a third prototype is missing. Below the first two prototypes that are present, just before the new paragraph beginning "There are three forms of...", there should be a third prototype. Please note that in the text below, detailed formatting is not exactly possible, so please arrange the formatting to match that of the two prototypes that are already present. That prototype should be:


void cv::calcHist(
cv::InputArrayOfArrays images, // vector of 8U or 32F images
const vector<int>& channels, // lists channels to use
cv::InputArray mask, // in 'images' count, iff 'mask' nonzero
cv::OutputArray hist, // output histogram array
const vector<int>& histSize, // hist sizes in each dimension
const vector<float>& ranges, // pairs give bin sizes in a flat list
bool accumulate = false // if true, add to 'hist', else replace
)

correct definition is

void cv::minMaxLoc ( const SparseMat & a,
double * minVal,
double * maxVal,
int * minIdx = 0,
int * maxIdx = 0
)

Note from the Author or Editor:
Reader is correct.

* Please change "cv::Point*" to "int*" in both places in which it appears in the function prototype just above the last paragraph on p381.

Data members nu20,... nu03 are normalized central moments rather than Hu invariant moments as is said in comments.

Note from the Author or Editor:
The reader is correct. Please make the change described below:

* Change the two lines reading:
"""
double nu20, nu11, nu02; // second order Hu invariant moments (x3)
double nu30, nu21, nu12, nu03; // third order Hu invariant moments (x4)
"""
to instead read:
"""
double nu20, nu11, nu02; // second order normalized central (x3)
double nu30, nu21, nu12, nu03; // third order normalized central (x4)
"""

The expression for the second Hu moment (h2) is incorrect - the plus should be changed for the minus.

Note from the Author or Editor:
Reader is correct, see change below:

Change "h_{2} = (\nu_{20} + \nu_{02})^{2}" to "h_{2} = (\nu_{20} - \nu_{02})^{2}". (i.e. Change the plus to a minus in the definition of h_2. at the top of page 434.)

int left_to_right = 0 // 1='fixed iteration direction'

should read as

bool left_to_right = 0 // true='always iteratу from left to right'

Note from the Author or Editor:
Reader is correct.

* Please change "int left_to_right = 0 // 1='fixed iteration direction'" to read "bool left_to_right = false // true='fixed iteration direction'".

The statement to open stdin

cap.open(0);

is redundant because main exits anyway.

Note from the Author or Editor:
Reader is correct. In this case, the call is unneeded. It is also somewhat confusing as written. It might have made more sense to call cap.close() in this location. However, because the cv:VideoCapture destructor will automatically call close(), the line can be completely removed.

* Delete the line reading "cap.open(0);"

updateCodebook() adds a new codebook...

should read

updateCodebook() adds a new codeword

Note from the Author or Editor:
Reader is correct.

* Please change "codebook if needed" to "codeword if needed" (about 3/4 of the way down p473)

The whole footnote should be deleted because, starting from version 3.0, class cv::BackgroundSubtractor is indeed abstract - all its member functions are pure virtual.

Note from the Author or Editor:
Reader is correct. This is a case in which the evolution of the library got ahead of the book.

* Delete footnote 18 completely from the bottom of page 485 (and superscript reference to footnote at the middle of the same page)

There are a few minor errors in the code listing for example 16-1:


In the definition of the help() function it currently reads:
void help( argv ) {

argv needs a type, e.g.:
void help( char **argv) {


Within the main() function, at the top of page 510 MAX_CORNERS is being redefined. It was previously defined near the very top of the file, following the "using" statement.


In the call to calcOpticalFlowPyrLK, the noArray() argument needs the cv:: namespace qualifier.


In the call to line() at the top of page 511, the last argument reads CV::AA, but it should be CV::LINE_AA.

Note from the Author or Editor:
Reader is correct, please make the following changes and corrections:

* Change "help( argv )" to "help( char** argv )" p509
* Change "noArray()" to "cv::noArray" (about 3/4 of the way down p510)
* replace line "const int MAX_CORENERS = 500;" with a blank line (top of p510)
* change "cv::imread argv[2]" with "cv::imread( argv[2]" (bottom of p509, note added parenthesis.)
* change "AA" to "LINE_AA" (top of p511)

Finally, the clone() and create() methods allow you to make a copy of a descriptor or create a new one by name, respectively.

should read

Finally, the clone() and create() methods allow you to make a copy of a matcher or create a new one by name, respectively.

Note from the Author or Editor:
Reader is correct, please change "of a descriptor or create" to "of a matcher or create" (middle of p525).

The sentence

Note that by default the locations of the features are not scaled, only the scale value assigned to them.

is the other way round. By default the step between features is scaled (bool varyXyStepWithScale = true)

Note from the Author or Editor:
Reader is correct

* please delete the sentence "Note that by default the locations of the features are not scaled, only the scale value assigned to them." (bottom of p570).

* also change "This will be the spacing between all of the features on all levels, unless varyXyStepWithScale is set to true, in which case the step between the features will also be scaled by one power of featureScaleMul for each level above the first that scale was generated on." to read:
"If featureScaleLevels is larger than one, then the locations of the features will, by default, also be scaled by one power of featureScaleMul for each level after the first. If varyXyStepWithScale is set to false, then only the scale value is assigned to the features will change (and not their locations)." (Please also note that in the above text: 'featureScaleLevels', 'featureScaleMul', 'varyXyStepWithScale', and 'false', should all be in "inline code" font.)

No, figures 16-2, 16-4, and 16-9 have nothing to do with drawKeypoints()

Note from the Author or Editor:
Reader is correct.

* Please change "Figures 16-2, 16-4, and 16-9" to "Figures 16-11, 16-13, and 16-17".

The argument count is an integer equal to the number of corners.

There's no such argument. The sentence should be removed.

Note from the Author or Editor:
Reader is correct.

* Please remove completely the sentence "The argument count is an integer equal to the number of corners." from the second to last paragraph on p 657