[GSoC] Add GIF decode and encode for imgcodecs by redhecker · Pull Request #25691 · opencv/opencv

redhecker · 2024-06-02T13:41:29Z

this is related to #24855

we add gif support for imread, imreadmulti, imwrite and imwritemulti

opencv_extra: opencv/opencv_extra#1203

Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

I agree to contribute to the project under Apache 2 License.
To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
The PR is proposed to the proper branch
There is a reference to the original bug report and related work
There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
The feature is well documented and sample code can be built with the project CMake

sturkmen72 · 2024-06-02T14:39:44Z

test_imgcodecs fails
probably you need to rename your repositories

https://github.com/redhecker/opencv-GSoC24 -> https://github.com/redhecker/opencv
https://github.com/redhecker/opencv_extra-GSoC24 -> https://github.com/redhecker/opencv_extra

redhecker · 2024-06-02T15:35:18Z

test_imgcodecs fails probably you need to rename your repositories

https://github.com/redhecker/opencv-GSoC24 -> https://github.com/redhecker/opencv
https://github.com/redhecker/opencv_extra-GSoC24 -> https://github.com/redhecker/opencv_extra

i've already done this, then how to rerun the build bot? or do i need to start a new pull request?

sturkmen72 · 2024-06-02T16:35:18Z

just add a new commit.

for example you can write some information here
https://github.com/redhecker/opencv/blob/4.x/modules/imgcodecs/include/opencv2/imgcodecs.hpp#L287

/** @brief Loads images from the specified file into a vector of Mat objects. 

The function imreadmulti loads images from the specified file into a vector of Mat objects.
if the file is one of supported multi-page file formats, all images will be loaded. Otherwise one image will be loaded.

Currently, the following file formats are supported:

-   GIF files - \*.gif (always supported)
-   AVIF - \*.avif (see the *Note* section)
-   TIFF files - \*.tiff, \*.tif (see the *Note* section)

@param filename Name of file to be loaded.
@param mats A vector of Mat objects holding each page.
@param flags Flag that can take values of cv::ImreadModes, default with cv::IMREAD_ANYCOLOR.
@sa cv::imread
*/

modules/imgcodecs/src/loadsave.cpp

vrabaud · 2024-06-03T13:38:01Z

You need to update the CMake code to take GIF into account. The best is probably to have a look at how it's done for AVIF.

vrabaud · 2024-06-03T14:26:27Z

Thx for the code. I'd recommend to wait for the animation API that @sturkmen72 will during the GSoC.
*multi is for multipage images, e.g. TIFF, and not animations.

modules/imgcodecs/src/grfmt_gif.cpp

modules/imgcodecs/test/test_gif.cpp

modules/imgcodecs/src/grfmt_gif.cpp

…r(0, 0, 0));

Co-authored-by: TrendMYX <trendmyx@126.com>

vrabaud

The biggest issue I see is with transparency: it seems not be handled properly. There should be an option to apply the background color or not (which would therefore return cv::Mat4b)

modules/imgcodecs/src/grfmt_gif.cpp

modules/imgcodecs/src/grfmt_gif.hpp

…e, imcount and use_rgb

Co-authored-by: TrendMYX <trendmyx@126.com>

modules/imgcodecs/src/grfmt_gif.cpp

asmorkalov · 2024-12-06T06:20:49Z

@vrabaud What is the current PR status? Is it ready for integration? Please list issues, if not.

CMakeLists.txt

sturkmen72 · 2025-01-08T07:44:21Z

@redhecker could you check this file https://upload.wikimedia.org/wikipedia/commons/d/d3/Newtons_cradle_animation_book_2.gif
imreadmulti and imreadanimation fails to load all frames

sturkmen72 · 2025-01-08T07:48:28Z

take a look at the source of file only the file above is problematic
https://commons.wikimedia.org/wiki/File:Newtons_cradle_animation_book_2.gif

redhecker · 2025-01-08T07:52:33Z

@redhecker could you check this file https://upload.wikimedia.org/wikipedia/commons/d/d3/Newtons_cradle_animation_book_2.gif imreadmulti and imreadanimation fails to load all frames

ok, i will try to fix this as soon as possible

sturkmen72 · 2025-01-08T08:34:00Z

in bool GifDecoder::getFrameCount_()

m_frame_count ++; reaches 9

imcount("d:/github/Newtons_cradle_animation_book_2.gif") returns 9

redhecker · 2025-01-08T12:35:53Z

in bool GifDecoder::getFrameCount_()

m_frame_count ++; reaches 9

imcount("d:/github/Newtons_cradle_animation_book_2.gif") returns 9

wait, when i run the following codes, i got 36 as output? i don't think i fully understand what you mean, do you mean you get 9 as output with the folling codes?

#include <opencv2/opencv.hpp>
#include <bits/stdc++.h>
int main() {
    std::cout << cv::imcount("Newtons_cradle_animation_book_2.gif") << std::endl;
}

but this cannot work for imreadmulti and imreadanimation. i think it's somewhere else that are wrong

sturkmen72 · 2025-01-08T13:25:40Z

in bool GifDecoder::getFrameCount_()
m_frame_count ++; reaches 9
imcount("d:/github/Newtons_cradle_animation_book_2.gif") returns 9

wait, when i run the following codes, i got 36 as output? i don't think i fully understand what you mean, do you mean you get 9 as output with the folling codes?
#include <opencv2/opencv.hpp>
#include <bits/stdc++.h>
int main() {
    std::cout << cv::imcount("Newtons_cradle_animation_book_2.gif") << std::endl;
}
but this cannot work for imreadmulti and imreadanimation. i think it's somewhere else that are wrong

sorry my bad.. you are right. i accidentally overwrite the image only readed frames when testing.

sturkmen72 · 2025-01-08T13:34:02Z

hasRead = lzwDecode(); // hasRead is false at 10th frame
https://github.com/opencv/opencv/blob/4.x/modules/imgcodecs/src/grfmt_gif.cpp#L149

redhecker · 2025-01-08T15:21:43Z

@sturkmen72 pls have a look at the newest pr, are there any other errors?

sturkmen72 · 2025-01-08T16:03:35Z

@redhecker well done! thank you.

Fix bugs in GIF decoding #26738 ### Pull Request Readiness Checklist this is related to #25691 i solved two bugs here: 1. the decoding setting: according to [https://www.w3.org/Graphics/GIF/spec-gif89a.txt](https://www.w3.org/Graphics/GIF/spec-gif89a.txt) ``` DEFERRED CLEAR CODE IN LZW COMPRESSION There has been confusion about where clear codes can be found in the data stream. As the specification says, they may appear at anytime. There is not a requirement to send a clear code when the string table is full. It is the encoder's decision as to when the table should be cleared. When the table is full, the encoder can chose to use the table as is, making no changes to it until the encoder chooses to clear it. The encoder during this time sends out codes that are of the maximum Code Size. As we can see from the above, when the decoder's table is full, it must not change the table until a clear code is received. The Code Size is that of the maximum Code Size. Processing other than this is done normally. Because of a large base of decoders that do not handle the decompression in this manner, we ask developers of GIF encoding software to NOT implement this feature until at least January 1991 and later if they see that their particular market is not ready for it. This will give developers of GIF decoding software time to implement this feature and to get it into the hands of their clients before the decoders start "breaking" on the new GIF's. It is not required that encoders change their software to take advantage of the deferred clear code, but it is for decoders. ``` at first i didn't consider this case, thus leads to a bug discussed in #25691. the changes made in function lzwDecode() is aiming at solving this. 2. the fetch method of loopCount: in the codes at https://github.com/opencv/opencv/blob/4.x/modules/imgcodecs/src/grfmt_gif.cpp#L410, if the branch is taken, 3 more bytes will be taken, leading to unpredictable behavior. See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request - [x] I agree to contribute to the project under Apache 2 License. - [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV - [x] The PR is proposed to the proper branch - [x] There is a reference to the original bug report and related work - [ ] There is accuracy test, performance test and test data in opencv_extra repository, if applicable Patch to opencv_extra has the same branch name. - [ ] The feature is well documented and sample code can be built with the project CMake

Fix bugs in GIF decoding opencv#26738 ### Pull Request Readiness Checklist this is related to opencv#25691 i solved two bugs here: 1. the decoding setting: according to [https://www.w3.org/Graphics/GIF/spec-gif89a.txt](https://www.w3.org/Graphics/GIF/spec-gif89a.txt) ``` DEFERRED CLEAR CODE IN LZW COMPRESSION There has been confusion about where clear codes can be found in the data stream. As the specification says, they may appear at anytime. There is not a requirement to send a clear code when the string table is full. It is the encoder's decision as to when the table should be cleared. When the table is full, the encoder can chose to use the table as is, making no changes to it until the encoder chooses to clear it. The encoder during this time sends out codes that are of the maximum Code Size. As we can see from the above, when the decoder's table is full, it must not change the table until a clear code is received. The Code Size is that of the maximum Code Size. Processing other than this is done normally. Because of a large base of decoders that do not handle the decompression in this manner, we ask developers of GIF encoding software to NOT implement this feature until at least January 1991 and later if they see that their particular market is not ready for it. This will give developers of GIF decoding software time to implement this feature and to get it into the hands of their clients before the decoders start "breaking" on the new GIF's. It is not required that encoders change their software to take advantage of the deferred clear code, but it is for decoders. ``` at first i didn't consider this case, thus leads to a bug discussed in opencv#25691. the changes made in function lzwDecode() is aiming at solving this. 2. the fetch method of loopCount: in the codes at https://github.com/opencv/opencv/blob/4.x/modules/imgcodecs/src/grfmt_gif.cpp#L410, if the branch is taken, 3 more bytes will be taken, leading to unpredictable behavior. See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request - [x] I agree to contribute to the project under Apache 2 License. - [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV - [x] The PR is proposed to the proper branch - [x] There is a reference to the original bug report and related work - [ ] There is accuracy test, performance test and test data in opencv_extra repository, if applicable Patch to opencv_extra has the same branch name. - [ ] The feature is well documented and sample code can be built with the project CMake

[GSoC] Add GIF decode and encode for imgcodecs opencv#25691 this is related to opencv#24855 we add gif support for `imread`, `imreadmulti`, `imwrite` and `imwritemulti` opencv_extra: opencv/opencv_extra#1203 ### Pull Request Readiness Checklist See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request - [x] I agree to contribute to the project under Apache 2 License. - [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV - [x] The PR is proposed to the proper branch - [x] There is a reference to the original bug report and related work - [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable Patch to opencv_extra has the same branch name. - [ ] The feature is well documented and sample code can be built with the project CMake

Fix bugs in GIF decoding opencv#26738 ### Pull Request Readiness Checklist this is related to opencv#25691 i solved two bugs here: 1. the decoding setting: according to [https://www.w3.org/Graphics/GIF/spec-gif89a.txt](https://www.w3.org/Graphics/GIF/spec-gif89a.txt) ``` DEFERRED CLEAR CODE IN LZW COMPRESSION There has been confusion about where clear codes can be found in the data stream. As the specification says, they may appear at anytime. There is not a requirement to send a clear code when the string table is full. It is the encoder's decision as to when the table should be cleared. When the table is full, the encoder can chose to use the table as is, making no changes to it until the encoder chooses to clear it. The encoder during this time sends out codes that are of the maximum Code Size. As we can see from the above, when the decoder's table is full, it must not change the table until a clear code is received. The Code Size is that of the maximum Code Size. Processing other than this is done normally. Because of a large base of decoders that do not handle the decompression in this manner, we ask developers of GIF encoding software to NOT implement this feature until at least January 1991 and later if they see that their particular market is not ready for it. This will give developers of GIF decoding software time to implement this feature and to get it into the hands of their clients before the decoders start "breaking" on the new GIF's. It is not required that encoders change their software to take advantage of the deferred clear code, but it is for decoders. ``` at first i didn't consider this case, thus leads to a bug discussed in opencv#25691. the changes made in function lzwDecode() is aiming at solving this. 2. the fetch method of loopCount: in the codes at https://github.com/opencv/opencv/blob/4.x/modules/imgcodecs/src/grfmt_gif.cpp#L410, if the branch is taken, 3 more bytes will be taken, leading to unpredictable behavior. See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request - [x] I agree to contribute to the project under Apache 2 License. - [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV - [x] The PR is proposed to the proper branch - [x] There is a reference to the original bug report and related work - [ ] There is accuracy test, performance test and test data in opencv_extra repository, if applicable Patch to opencv_extra has the same branch name. - [ ] The feature is well documented and sample code can be built with the project CMake

sturkmen72 · 2025-07-18T12:00:02Z

modules/imgcodecs/test/test_gif.cpp

+    param.push_back(7);
+    param.push_back(IMWRITE_GIF_DITHER);
+    param.push_back(2);
+    EXPECT_NO_THROW(imencode(".png", img_gt, buff, param));


@redhecker is there something like forgotten issue here ?
EXPECT_NO_THROW(imencode(".gif", img_gt, buff, param)); fails

I think that there's something wrong here, where ".png" used with gif params. However, the code was written by @TrendMYX , i think he will fix it in a few days.

@redhecker @TrendMYX fyi

[ERROR:0@0.088] global loadsave.cpp:1687 cv::imencodeWithMetadata imencode(): can't encode data: OpenCV(4.13.0-dev) C:\build\precommit_windows64\4.x\opencv\modules\imgcodecs\src\grfmt_gif.cpp:1022: error: (-215:Assertion failed) false in function 'cv::GifEncoder::ditheringKernel' Exception message: OpenCV(4.13.0-dev) C:\build\precommit_windows64\4.x\opencv\modules\imgcodecs\src\loadsave.cpp:1274: error: (-215:Assertion failed) !buf.empty() in function 'cv::imdecode_'

@sturkmen72 @redhecker Sorry for all the mistakes. Here is the modified version which goes through successfully on my laptop.

TEST(Imgcodecs_Gif,write_gif_flags){ //rows611 x cols293, three‑channel, constant value 1 const int expected_rows = 611; const int expected_cols = 293; cv::Mat img_gt(expected_rows, expected_cols, CV_8UC3, cv::Scalar::all(1)); //the encoding parameters std::vector<unsigned char> buff; std::vector<int> param; param.push_back(cv::IMWRITE_GIF_QUALITY); param.push_back(7); param.push_back(cv::IMWRITE_GIF_DITHER); param.push_back(2); //Encode part ASSERT_NO_THROW(cv::imencode(".gif", img_gt, buff, param)); //Decode part cv::Mat img; ASSERT_NO_THROW(img = cv::imdecode(buff, cv::IMREAD_ANYCOLOR)); ASSERT_FALSE(img.empty()); EXPECT_EQ(img.cols, expected_cols); EXPECT_EQ(img.rows, expected_rows); EXPECT_EQ(img.type(), CV_8UC3); // Compare, maxdiff=16 EXPECT_PRED_FORMAT2(cvtest::MatComparator(16, 0), img, img_gt); }

@TrendMYX thank you. feel free to create a PR or i will do later.

redhecker and others added 2 commits June 2, 2024 19:59

enable gif decode

3ab9937

add gif_test.cpp (#1)

1c80789

redhecker mentioned this pull request Jun 2, 2024

add gif test data opencv/opencv_extra#1182

Closed

opencv-alalek added the GSoC label Jun 2, 2024

add explicit cast and modified test_gif to avoid warnings

91b8507

asmorkalov requested a review from vrabaud June 3, 2024 09:12

asmorkalov changed the title ~~[GSoC]Add GIF decode for imgcodecs~~ [GSoC] Add GIF decode for imgcodecs Jun 3, 2024

asmorkalov added feature category: imgcodecs labels Jun 3, 2024

asmorkalov added this to the 4.11.0 milestone Jun 3, 2024

vrabaud reviewed Jun 3, 2024

View reviewed changes

modules/imgcodecs/src/loadsave.cpp Outdated Show resolved Hide resolved

asmorkalov reviewed Jun 10, 2024

View reviewed changes

redhecker added 6 commits June 12, 2024 00:14

add cmake configuration

2ce2453

check signature

9acaead

use cv::Mat::zeros() instead of Mat(m_height, m_width, CV_8UC3, Scala…

8a8782c

…r(0, 0, 0));

use vector instead of array

8a5dfd7

modified gif test

327c271

Co-authored-by: TrendMYX <trendmyx@126.com>

correct typo

ef1f471

redhecker requested review from asmorkalov and vrabaud July 6, 2024 01:01

vrabaud reviewed Jul 15, 2024

View reviewed changes

modules/imgcodecs/src/grfmt_gif.cpp Outdated Show resolved Hide resolved

modules/imgcodecs/src/grfmt_gif.hpp Outdated Show resolved Hide resolved

redhecker and others added 4 commits July 27, 2024 18:33

Merge branch 'opencv:4.x' into gifSupport

f61b179

modified gif decode: add support for transparent color, no color tabl…

bcfeea9

…e, imcount and use_rgb

add test for new feature

2662d0c

Co-authored-by: TrendMYX <trendmyx@126.com>

modified test data

15e4b68

Co-authored-by: TrendMYX <trendmyx@126.com>

redhecker changed the title ~~[GSoC] Add GIF decode for imgcodecs~~ [GSoC] Add GIF decode and encode for imgcodecs Aug 21, 2024

redhecker and others added 2 commits August 27, 2024 17:43

prevent overflow when shift left

9a6992b

add accuracy tests for gif encoder

52c88bc

Co-authored-by: TrendMYX <trendmyx@126.com>

dkurt reviewed Sep 12, 2024

View reviewed changes

modules/imgcodecs/src/grfmt_gif.cpp Outdated Show resolved Hide resolved

modules/imgcodecs/src/grfmt_gif.cpp Outdated Show resolved Hide resolved

add fast mode for encoder by avoid octree color quantization

6ea52dd

Code review fixes.

d767145

mshabunin reviewed Dec 6, 2024

View reviewed changes

CMakeLists.txt Outdated Show resolved Hide resolved

Disabled GIF by default for now.

8b1fe27

asmorkalov approved these changes Dec 7, 2024

View reviewed changes

asmorkalov merged commit 082cd7a into opencv:4.x Dec 7, 2024

redhecker mentioned this pull request Jan 8, 2025

fix bugs in GIF decoding #26738

Merged

6 tasks

mshabunin mentioned this pull request Jan 9, 2025

opencv: build with GIF support google/oss-fuzz#12927

Merged

asmorkalov mentioned this pull request Jan 15, 2025

5.x merge 4.x #26768

Merged

sturkmen72 reviewed Jul 18, 2025

View reviewed changes

Uh oh!

Comments

Conversation

redhecker commented Jun 2, 2024 • edited by asmorkalov Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Pull Request Readiness Checklist

Uh oh!

sturkmen72 commented Jun 2, 2024

Uh oh!

redhecker commented Jun 2, 2024

Uh oh!

sturkmen72 commented Jun 2, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

vrabaud commented Jun 3, 2024

Uh oh!

vrabaud commented Jun 3, 2024

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

vrabaud left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

asmorkalov commented Dec 6, 2024

Uh oh!

Uh oh!

sturkmen72 commented Jan 8, 2025

Uh oh!

sturkmen72 commented Jan 8, 2025

Uh oh!

redhecker commented Jan 8, 2025

Uh oh!

sturkmen72 commented Jan 8, 2025

Uh oh!

redhecker commented Jan 8, 2025

Uh oh!

sturkmen72 commented Jan 8, 2025

Uh oh!

sturkmen72 commented Jan 8, 2025

Uh oh!

redhecker commented Jan 8, 2025

Uh oh!

sturkmen72 commented Jan 8, 2025

Uh oh!

sturkmen72 Jul 18, 2025

Choose a reason for hiding this comment

Uh oh!

redhecker Jul 18, 2025

Choose a reason for hiding this comment

Uh oh!

sturkmen72 Jul 18, 2025

Choose a reason for hiding this comment

Uh oh!

TrendMYX Jul 18, 2025

Choose a reason for hiding this comment

Uh oh!

sturkmen72 Jul 19, 2025

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

redhecker commented Jun 2, 2024 •

edited by asmorkalov

Loading

sturkmen72 commented Jun 2, 2024 •

edited

Loading