new module: intensity_transform #2399

elim-dk · 2019-12-31T06:32:59Z

The new module, intensity_transform, implements the following intensity transformation algorithms: gamma correction, log transformation, autoscaling, and contrast stretching.

This pullrequest changes

Adds a new module called intensity_transform.

alalek

Thank you for the contribution!

Please take a look on comments below.

alalek · 2020-01-09T06:15:23Z

modules/intensity_transform/CMakeLists.txt

@@ -0,0 +1,2 @@
+set(the_description "Intensity transformations")
+ocv_define_module(intensity_transform opencv_core opencv_imgcodecs WRAP python)


imgcodecs

Looks like this dependency is not used in module itself ("samples" has own extra dependencies list).
Please remove this from here and precomp.hpp

alalek · 2020-01-09T06:18:50Z

modules/intensity_transform/include/opencv2/intensity_transform.hpp

+ * @param input input bgr or grayscale image.
+ * @param output resulting image of log transformations.
+*/
+CV_EXPORTS_W void logTransform(const cv::Mat input, cv::Mat& output);


[code style] Please avoid using of "cv::" in OpenCV files for Mat.

alalek · 2020-01-09T06:21:26Z

modules/intensity_transform/src/intensity_transform.cpp

+
+void contrastStretching(const Mat input, Mat& output, const int r1, const int s1, const int r2, const int s2)
+{
+    vector<int> table(256);


std::array ?

Thank you for your comment. Since cv::LUT method takes in cv::_InputArray as the argument, std::array didn't work here.

It should work, related code is here.

BTW, Use std::array<uchar, 256> table; and remove final .convertTo().

Thanks for the document! But, when I use std::array<uchar, 256> table;, I get the following error:
candidate function not viable: no known conversion from 'std::array<uchar, 256>' (aka 'array<unsigned char, 256>') to 'const cv::_InputArray' for 2nd argument CV_EXPORTS_W void LUT(InputArray src, InputArray lut, OutputArray dst);
Any idea why?

Which compiler is used?
Which version of code from main "opencv" repository is used?

Could you try LUT(..., InputArray(table), ...)?

This commit works well (pass CI checks, see #2366): b05b328

Great, thank you! Made the change.

alalek · 2020-01-09T06:25:23Z

modules/intensity_transform/test/test_intensity_transform.cpp

+
+TEST(intensity_transform_logTransform, accuracy)
+{
+    cv::Mat const image = (cv::Mat_<uchar>(10, 10) <<


Could you please avoid using of this initializer in OpenCV? Consider using RAW array instead.
see opencv/opencv#15959 (comment)

alalek · 2020-01-09T06:29:08Z

modules/intensity_transform/test/test_intensity_transform.cpp

+    {
+        for(int j = 0; j < res.rows; ++j)
+        {
+            EXPECT_EQ((int)res.at<uint8_t>(i,j), (int)expectedRes.at<uint8_t>(i,j));


loops =>

EXPECT_EQ(0, cvtest::norm(res, expectedRes, NORM_INF));

Please note that this check is very strong (and should be avoided due non bit-exact floating-point implementation).
It is better to relax conditions:

EXPECT_LE(cvtest::norm(res, expectedRes, NORM_INF), 1); EXPECT_LE(cvtest::norm(res, expectedRes, NORM_L1), 10);

alalek · 2020-01-20T12:33:50Z

modules/intensity_transform/src/intensity_transform.cpp

+    if (input.channels() == 1)
+    {
+        output = input + 1;
+    }
+    else
+    {
+        output = input + Scalar(1, 1, 1);
+    }
+
+    output.convertTo(output, CV_64F);
+    cv::log(output, output);
+    output *= c;
+    output.convertTo(output, CV_8UC3);


Mat::convertTo() with alpha/beta parameters can reduce number of operations:

Mat in_plus_1_64f; input.convertTo(in_plus_1_64f, CV_64F, 1, 1.0f); // input + 1 Mat log_64f; log(in_plus_1_64f, log_64f); log_64f.convertTo(output, CV_8UC3, c, 0.0f); // log_64f * c

That's useful! Thank you for the feedback!

alalek · 2020-01-20T12:45:43Z

modules/intensity_transform/src/intensity_transform.cpp

+
+void contrastStretching(const Mat input, Mat& output, const int r1, const int s1, const int r2, const int s2)
+{
+    vector<int> table(256);


It should work, related code is here.

BTW, Use std::array<uchar, 256> table; and remove final .convertTo().

…fomration, autoscaling, contrast stretching) build error fix: remove trailing whitespaces, casting types minor edits based on alalek's feedback removing trailing whitespace using std::array for LUT argument

alalek

Well done! Thank you 👍

…ions

elim-dk · 2020-01-29T18:50:55Z

Well done! Thank you 👍

Thank you for all your constructive comments and help!

elim-dk force-pushed the algorithm/intensity-transformations branch from 5f74f07 to bf25db3 Compare December 31, 2019 18:12

alalek reviewed Jan 9, 2020

View reviewed changes

elim-dk force-pushed the algorithm/intensity-transformations branch 2 times, most recently from 9383248 to 88faaa7 Compare January 12, 2020 15:48

alalek reviewed Jan 20, 2020

View reviewed changes

new module: intensity_transform (includes gamma correction, log trans…

9aa74df

…fomration, autoscaling, contrast stretching) build error fix: remove trailing whitespaces, casting types minor edits based on alalek's feedback removing trailing whitespace using std::array for LUT argument

elim-dk force-pushed the algorithm/intensity-transformations branch from 88faaa7 to 9aa74df Compare January 27, 2020 21:44

alalek approved these changes Jan 29, 2020

View reviewed changes

opencv-pushbot pushed a commit that referenced this pull request Jan 29, 2020

Merge pull request #2399 from elim-dk:algorithm/intensity-transformat…

2dcd366

…ions

opencv-pushbot merged commit 9aa74df into opencv:master Jan 29, 2020

		@@ -0,0 +1,2 @@
		set(the_description "Intensity transformations")
		ocv_define_module(intensity_transform opencv_core opencv_imgcodecs WRAP python)

new module: intensity_transform #2399

new module: intensity_transform #2399

Uh oh!

Conversation

elim-dk commented Dec 31, 2019

This pullrequest changes

Uh oh!

alalek left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

alalek left a comment

Choose a reason for hiding this comment

Uh oh!

elim-dk commented Jan 29, 2020

Uh oh!

Uh oh!