Blurring and Smoothing - OpenCV with Python for Image and Video Analysis 8

Key Takeaways

what's going on everybody welcome to another open CV with Python tutorial video where we left off we were filtering for a specific color you can imagine there's a lot of different filters that you might apply may not even be for a specific color where you apply the filter but you still have a lot of background noise and later on when we do like foreground extraction or background extraction the same thing happens you wind up still with a little bit of noise around just things are moving or whatever falsely detected and so on so you need to have some way to be able to remove noise so first we'll be talking about really simplistic ways like smoothing and blurring to get rid of noise and then we'll continue talking about in the subsequent tutorial how to even further get rid of noise in different types of noise so so anyways let's go ahead get started let me grab my head again and so right now we have the following we've got just the the regular frame the mask in the resolution we don't really or the result we don't need to we could probably keep masks just because it's easy to see the background stuff actually but we'll probably just keep doing the result instead and we'll stop displaying masks but anyway that's what we're dealing with so far so let's say you want to just apply like a really simple blur basically and that's you can basically do this with averaging the pixels so um I said we'd stop label set up a display masks because we got so many things for displaying feel free to like only display frame or only to spray the display the one thing we're working on like if you're having trouble with lag or something like that so to do averaging it's relatively simple we're going to make a kernel and the kernel will be numpy ones just a list of ones basically 15 by 15 datatype numpy float 32 and it's going to be so it's 15 by 15 and then we divide it by whatever this number times this number is in RK is 15 times 15 - 25 so that I make an average of 15 by 15 pixels okay you can go smaller you can say five by five and that would be 25 okay so that's your kernel and then we apply that kernel by saying smooth equals CV to filter 2d and then we're gonna say res negative one and calm so this will just apply like a simple average basically of our frame oh we need to display it almost made the same mistake as before okay so pay space good to go all right so we actually I mean with this we basically have no background noise but I see a few little specks back there but obviously it gets rid of a lot but we also lose a lot of clarity so it's kind of a bummer so we close this and we we assign and that's not really the best right and that's okay so now we're going to talk about another form and that will be a Gaussian blur so we'll just do it right underneath here and we're going to say blur blur blur is equal to CB - Gaussian blur we're going to apply it to that results this - will be a 15 by 15 in n0 and then we need to show it so let's copy that paste and blur and blur that and it even though it's like the same kind of size you should see this one's a lot clearer more clear than the other blur or average basically was so there's that option my favorite though for clarity is what's called this median blur so median will be equal to cv - median blur and we apply this to the results and it's a 15 of size so then we're going to call this paced medium meaning run that and at least here especially at a distance like if you backup and this thing is small like smaller on the frame you should notice that it almost like we can make it disappear and and I guess I mean that's just what's happening as opposed as you move it back it just doesn't want to show up I guess it's like falling out of the range that I specified probably but if we hold it like nice and close this one is the best as far as internal noise is concerned there's really no noise happening whereas there's a lot of noise happening over there and it's of course being returned in that blur but I happen to like this one the best just simply because it's the least noisy so anyway there's that and then finally there is a bilateral blur it's just another version of blurring so bilateral equals CV to bilateral filter apply to the res 15 then we'll choose a 75 by 75 and then see what call CB TM show down here just for the different versions of blurs okay so in my opinion the bilateral blur really is pretty much useless it got rid of a little bit of noise but not really so okay so anyways those are the various blurs that you have and I suppose a smooth that you have at your disposal but a lot of times that's just simply not going to be enough so in the next tutorial we're going to be talking about as morphological transformations and what that's going to allow us to do is is remove noise and a bit better of a method so anyway that's we're gonna be covering this tutorial if you have questions comments concerns or whatever up to this point feel free to let me know below otherwise as always thanks for watching thanks for all the support subscriptions in its seal next time