That's what I was thinking ,too !! Much lesser time than others which is incredibly needed for us as the most people don't wanna spend more than 15 ,or 10 minutes to watch videos on phone !! Some are super busy . 👍👍💜🥁🐉🎤🎶💞
Though I couldn't understand the complete explanation due to my lack of the knowledge on this particular subject, I got how it relatively woks :) Thank you for this video!
duuude amazing content. My class covered this topic in like 2 seasons, 2h in total and you broke it down to 7 min. Well, to be fair it is also more detailed than this but nice overview of basic principles
you are so much great my professor only teach the compression technique but dont even explain when and how and why they are used m thanks to you i understand this now
The information about frequency dependant contrast sensitivity is way too interesting. Would you please link more resources? I am mostly interested in the fact that it varies from person to person and as my curve peak is moved quite a bit to the right, I would want to know why.
It's indeed interesting. I've seen it in an image processing class taught by my doctoral advisor Alan Bovik. If I remember correctly, it was also covered in his book titled "The Essential Guide to Image Processing." You can also do a web search on the Contrast Sensitivity Function find more information about it.
very good explanation, will watch again to understand all the points. please look at fractal compression because it is as good at least for compression rate but have no information loss
your illustrations are very relatable...your voice is so clear...thank you for the videos....can you do a video on different image formats and different colour spaces?
Well the more those blocks getting quantized/blurred based on the frequenccy’s to remove, the more more blurrier the image will be,you could use sharpening to compensate for that but still.
Interesting. thumbs up. So in 6:50 you have 8x8 (=64 cells) which values can be from 0 to 255 in color range for each layer of color. So if you have every possible combination besides zig zag the Permutation total would be :o over 130 digits long. Again that is just for a 8x8 with 255 numbers for one color in total. What if your compression program was 1 gigabyte then would that mean the compressed file would be smaller since the program will have all the combinations sorted from highest to low. In the end its compression /speed ratio. but it all seems to come down to luck if the numbers show up just right. for example its easier to compress a number like this 3,486,784,401 (10 digits) to divided it by 9 at 9 times to be 1 digit (3 digits long total) 66% shrink down. Even at higher digits it becomes even more efficient over 80%. But to even remove 1 random digit from the number the math is off and good luck getting even 1/2 the efficiency. multiple techniques have to be used which one key thing many are not using which would help the most is to have the file be set up so that it can be knocked down to be compressed. Again this 10 digit 3,486,784,401 is faster and better to compress than this 5 digit 84,401. Better to have a bigger file structured properly than to have a smaller disorderly file. But then again who pays attention.
It’s possible to get a better compression than jpeg while preserving the perceptual image quality using more computationally expensive methods. I never tried using fractal image compression but it seems to rely on self similarly. Searching for similar image patches can require a lot of computation. It’s very hard to beat well established image compression methods without increasing computational complexity.
I really enjoyed this video. Thank you so much ! Just what is the effect of compression on noisy image (for example an image with gaussian noise). Thanks
Sure! Noise usually makes images harder to compress. A noisy image would have a larger size and lower quality after compression as compared to a clean image.
@@leoisikdogan Do you happen to know why early jpgs had very poor quality whites? In so many jpgs back in the 90's the color white nearly always came out blueish or dingy. I just saw it again on a video game from 1995ish... it was the Sony Interactive logo and all the whites were very off color. Just curious if you knew. I can't find anyone else talking about that.
I forgot why I wanted to watch the video or why it was open in the first place. But well, I learned a lot and it was a very nice video.... win-win I guess :D
Wow, that's well explained. The only thing I think is (kinda) wrong: It should not be Megabyte (MB), but instead Mebibyte (MiB). Because Megabytes are base 2 which means 12MB = 12.582.912 Bytes, and Mebibyte are base 10 which means 12MiB = 12.000.000 Bytes. Another fix would be to say that the original image is 4.096x3.072 pixels... However, this mistake is in almost every Literature, so it doesn't actually matter and, moreover, it's not the point of the video... ❤
Hi ! Thank you so much for your help! I was wondering, how did you manage to separate the Y, Cb and Cr images ? I have been searching all ver the internet but I can't find it.
Hi! If you are using OpenCV in Python, then you can do so by: img = cv2.cvtColor(img, cv2.COLOR_BGR2YCR_CB) Y = img[..., 0] Cr = img[..., 1] Cb = img[..., 2]
@@leoisikdogan Thanks ! But I don't know openCV unfortunately. I guess I was hopping for a easier photoshop solution. But I mean, you are capable of creating paintings with AI so ... this might explain why it's also hard to create this kind of pictures !
You can do it in Photoshop too. You can find them in the channels window next to the layers. You may need to change the color space from RGB to Lab first.
I am doing a project for memristor based hardware accelerator for image compression .In this project i using xilings ise design software and matlab .But in xilings ise desing 14.5 how to change when no of inputs given and change speed , area, delay plzzz tell .e
Hi, when scaling down images, sharpness is lost. A large image with 25% quality setting is sharper than a small image with 75% quality setting. However, I'm afraid that my google pagespeed score and seo rating will down (as pagespeed prefers correctly sized images)
![How are Images Compressed? [46MB ↘↘ 4.07MB] JPEG In Depth](http://i.ytimg.com/vi/Kv1Hiv3ox8I/mqdefault.jpg)
You have explained this in less than 10 minutes better than my professor for Image Processing in 90 minutes. Thanks.
That's what I was thinking ,too !! Much lesser time than others which is incredibly needed for us as the most people don't wanna spend more than 15 ,or 10 minutes to watch videos on phone !! Some are super busy . 👍👍💜🥁🐉🎤🎶💞
a 2 hours course compressed in a 6 minutes video. well done!
thank you so much I spent days trying to understand the compression steps and you illustrated them in minutes, I am so grateful
Thank you! You've shrinked my three uni lectures into 7 minute video!
The most easy to understand video I've found so far...after several days of struggling
And this was my entire semester, this is way better though
As a web dev completely new to image compression, I found this video a super helpful introduction, thank you!
Though I couldn't understand the complete explanation due to my lack of the knowledge on this particular subject, I got how it relatively woks :) Thank you for this video!
Oh my god someone finally explained it in a way that didn’t make my brain explode or fall asleep you have saved me
U just made it very easy for most of everyone to understand image compression
Thanks a lot man
More useful than I expected from the thumbnail 😅😅😅 I clicked this one since it is relatively a shorter one
Nicely done! Had a couple of 'a-haa' moments while watching.
Emre Arıkan Thanks! I'm glad to hear that.
@@leoisikdogan Dude, you are a real teacher. A real gem, above many others. Two videos in and my mind is blown again and again. Thank you!
duuude amazing content. My class covered this topic in like 2 seasons, 2h in total and you broke it down to 7 min. Well, to be fair it is also more detailed than this but nice overview of basic principles
RUclips should crown this human!
The first video on jpeg compression that actually explains why DCT is applied to the image
i have'nt expected that voice tone
I really enjoyed this video. I recently have delved into image processing for my work and this game me as many questions as answers. Great Stuff!
One of the best presentations I have ever seen
you are so much great my professor only teach the compression technique but dont even explain when and how and why they are used m thanks to you i understand this now
Read about this in my Image Processing class, but tbh we didn't had much time and clarity to this topic. Thanks !
I learnt more than
my 30 lectures of my semester
you flew over the frequency/quantization stuff in no time and I finally understood! thanks!
Amazing quick explanation. Thanks for the video!
You've done such a great job, man!.
Wow man, this is badass and a very thorough explanation! Not sure how this vid doesn't have a million views.
You deserve more subs. Never thought I'd be learning so deeply about a compressed imagine for my blog. Lol. Thanks!
Well explained! The way you illustrate the concepts is amazing.
You sir, are an absolute G! This was so clear!
Very nice explaining, so brief, useful, and understandable, thank you so much
واضح يعطيك العافية thank you it was very clear
The information about frequency dependant contrast sensitivity is way too interesting. Would you please link more resources? I am mostly interested in the fact that it varies from person to person and as my curve peak is moved quite a bit to the right, I would want to know why.
It's indeed interesting. I've seen it in an image processing class taught by my doctoral advisor Alan Bovik. If I remember correctly, it was also covered in his book titled "The Essential Guide to Image Processing." You can also do a web search on the Contrast Sensitivity Function find more information about it.
very good explanation, will watch again to understand all the points.
please look at fractal compression because it is as good at least for compression rate but have no information loss
Thank you for making it easy and understandable.
your illustrations are very relatable...your voice is so clear...thank you for the videos....can you do a video on different image formats and different colour spaces?
Good information, This video was explained well. Thank you
Hi, thanks for making this great video! May I ask what does each individual "value" in the 8x8 matrix mean in 3:47? Thank you!
I am doin this as my college project. Thanks a lot from India
Very informative!! and deserves more than a million views.
OMG this saves my final exam! thank you sooo much!!!
You just saved my math essay! Thanks a lot!
That was such a great video on a topic that I didn't know was even interesting before now! Thank you for that!
wow great detailing .. please which software do you recomend for image compresions
Very clear explanation~ Thank you
Woah ! This was an amazing watch!
Well the more those blocks getting quantized/blurred based on the frequenccy’s to remove, the more more blurrier the image will be,you could use sharpening to compensate for that but still.
This saved my life, thank you
Interesting. thumbs up. So in 6:50 you have 8x8 (=64 cells) which values can be from 0 to 255 in color range for each layer of color. So if you have every possible combination besides zig zag the Permutation total would be :o over 130 digits long. Again that is just for a 8x8 with 255 numbers for one color in total. What if your compression program was 1 gigabyte then would that mean the compressed file would be smaller since the program will have all the combinations sorted from highest to low. In the end its compression /speed ratio.
but it all seems to come down to luck if the numbers show up just right. for example its easier to compress a number like this 3,486,784,401
(10 digits) to divided it by 9 at 9 times to be 1 digit (3 digits long total) 66% shrink down. Even at higher digits it becomes even more efficient over 80%. But to even remove 1 random digit from the number the math is off and good luck getting even 1/2 the efficiency.
multiple techniques have to be used which one key thing many are not using which would help the most is to have the file be set up so that it can be knocked down to be compressed. Again this 10 digit 3,486,784,401 is faster and better to compress than this 5 digit 84,401. Better to have a bigger file structured properly than to have a smaller disorderly file. But then again who pays attention.
Nicely done! Can you make a video for dicom image compression and decompression?
really nice video. Hello from Kazakhstan
Fantastic breakdown. Love this.
Perfectly explained!!!
Thank you so much for this explanation.
This is a very useful video, thank you!
nicely done Leo!
Nice explanation Madam/sir thanks
Thank you for your wonderful explanation.
Very well explained! Thanks!
i got it perfect idea from this video. Thanks dude
Sir it is very nice video.But i have a doubt the fractal compression of images.How can it possible?
It’s possible to get a better compression than jpeg while preserving the perceptual image quality using more computationally expensive methods. I never tried using fractal image compression but it seems to rely on self similarly. Searching for similar image patches can require a lot of computation. It’s very hard to beat well established image compression methods without increasing computational complexity.
@@leoisikdogan 💓💓
Excellent explanation
Awesome video! You deserves 100M+ Views
I really enjoyed this video. Thank you so much ! Just what is the effect of compression on noisy image (for example an image with gaussian noise). Thanks
Sure! Noise usually makes images harder to compress. A noisy image would have a larger size and lower quality after compression as compared to a clean image.
@@leoisikdogan Do you happen to know why early jpgs had very poor quality whites? In so many jpgs back in the 90's the color white nearly always came out blueish or dingy. I just saw it again on a video game from 1995ish... it was the Sony Interactive logo and all the whites were very off color. Just curious if you knew. I can't find anyone else talking about that.
What you mean by subsampling, can you explain it, please? Thanks very much Leo!
I forgot why I wanted to watch the video or why it was open in the first place. But well, I learned a lot and it was a very nice video.... win-win I guess :D
Awesome Explanation
Amazing video. Thank you for this. I am inspired.
Thanks, happy to hear that!
Thank you Peter Parker, that was an extremely well put explanation.
Amazing explanetion!
nice explanation and thanks for shortening my lectures to little video, but let me be little bit critical, your voice
Wow, that's well explained. The only thing I think is (kinda) wrong: It should not be Megabyte (MB), but instead Mebibyte (MiB). Because Megabytes are base 2 which means 12MB = 12.582.912 Bytes, and Mebibyte are base 10 which means 12MiB = 12.000.000 Bytes.
Another fix would be to say that the original image is 4.096x3.072 pixels...
However, this mistake is in almost every Literature, so it doesn't actually matter and, moreover, it's not the point of the video... ❤
YOU SAVED MY LIFE
Hi ! Thank you so much for your help! I was wondering, how did you manage to separate the Y, Cb and Cr images ? I have been searching all ver the internet but I can't find it.
Hi! If you are using OpenCV in Python, then you can do so by:
img = cv2.cvtColor(img, cv2.COLOR_BGR2YCR_CB)
Y = img[..., 0]
Cr = img[..., 1]
Cb = img[..., 2]
@@leoisikdogan Thanks ! But I don't know openCV unfortunately. I guess I was hopping for a easier photoshop solution. But I mean, you are capable of creating paintings with AI so ... this might explain why it's also hard to create this kind of pictures !
You can do it in Photoshop too. You can find them in the channels window next to the layers. You may need to change the color space from RGB to Lab first.
@@leoisikdogan Yes it worked, thank you !
After huffman coding, is there any binary encoding ? Because Computers store memory in binary form.
very good explanation
you are amazing ..God bless you!
vector is beautiful it's like midi in images
I am doing a project for memristor based hardware accelerator for image compression .In this project i using xilings ise design software and matlab .But in xilings ise desing 14.5 how to change when no of inputs given and change speed , area, delay plzzz tell .e
Beautiful video
Hi, Thank you for video. Can you share a slide?
2:58 : the highlighted circle is high contrast high frequency region. Why you call it as "low contrast low high frequency"? Its confusing me.
Hi, when scaling down images, sharpness is lost.
A large image with 25% quality setting is sharper than a small image with 75% quality setting.
However, I'm afraid that my google pagespeed score and seo rating will down (as pagespeed prefers correctly sized images)
Can you build an algorithm to compress an image? It will be more interesting 🙂
wow...learned a lot...thanks Leo
Easy to understand
Hi, what a great video ! Thank you ,
Can you please share more about computer vision and image processing ?
Thank you
Thanks, sure! You can take a look at this one if you haven't seen it already: ruclips.net/video/9-8Js62wzQs/видео.html
@@leoisikdogan actually, it is what made me want to learn it efficiently!
@@bouchrad.339 Happy to hear that!
now its clear thank's alot
Nice work!
Hi, how to make image compression for live web camera ? Is there a resource for this ? github or videos ? can you help me ? please...
A very good video ! Thanks ! 👍👍💜🥁🐉🎤🎶💞
Excellent!
Not sure how I ended up here at 1am but like, not complaining haha
Thank you Leo, very cool
Really like your way ,, keep going man :D
Nicely explained video 🔥🔥🔥🔥
How can we do programmatically????
thank you for your video
good explanation, good content!
okay, someone does it with more technical terms!! :) well done
This is great!
Thanks a lot very helpful