CSCI 5722: Computer Vision | Spring 25 | Lecture 1 - Image | Highlights and Annotations by Gistr.

just an example 43, just five two five six seven. and there's a array of numbers that represent image and so then you have to define this, okay, this represent an image and then for simple image is it's two image is actually 2d thing to you a to convert array first you want to convert into matrix and the information we need to convert a rate into image matrix is the width, the image width in this case, image wid is two.03:08So with that information, I'll be able to say, okay, every time I have a row of pixels four and three, I'll wrap the row around after the width of two. And I have a second row two and five, my third row is six and seven.03:23So I have this image match representation. And if we have a different width information, for instance, three, then repeat the similar W process by a different parameter. In this case, the width is three. So I can do it by hand would be three, four, 3, two like this trying to start small. Okay so and the image is that instead of in just integers or some real numbers we like to have value between zero and one. So usually a binary image result of applying sort of a condition or filter.04:12So in this case what if we look at all the pixels 432. 5, six 7 but only want to see okay greater than three. And so the first location four yes, is greater than three. Three is not greater than three.04:24We get 0. 2. We get 0. 5, 6. 7. How about 5. 67? it's all 1, one one greater than three. So we get a b binary image like this. Another type is uh instead of, uh, just arbitrary numbers I just write some random numbers 432 But then if a bite array now SPE specify particular data type if a bite, how many bits are there in a bite is eight bits and what would be the number? How how, what's a like a range of values So a range for a bite in this case it could be from zero to two to the x power is 255. So double can be negative and be positive.06:24Also can be really small, Can be really large. So this is minus, uh, 3.4. e to the 308 power and max is power plus 3.4 If we have about we allow positive And negative numbers. So this is not something you need to remember but it's more about in practice With double array you'll be able to put number like 0. And another common thing we do is we wanted like to nor normalize this double array When we normalize a double array.07:50Our goal is that instead of some really large range like this OB large range. and we just have want to limit to zero and between zero and one. And this quite commonly use, especially when we get to a probability map as image.08:07which is all probability values and would between zero and one. So for instance we can have zero in our normalized double array. could have 0.4 and 0.3, 0.2 but nothing negative. It's 0.1, 0.3. This is the example of normalized double array.08:26So you can wrap them into a matrix. Okay now next I want to talk about some colors. you might have probably studied this or you've uh, develop any website working with CSS. you know this but this is more like a review. So if we have a bite and bite just a reminder zer to 255.08:52right? How do we get a red red color? So we have three different channels red with which is R GB. green and blue for red. we're going to have a highest intensity for the red channel. This are one of the three primary colors.09:08And so we have z z for the other one. So for green is obviously this one for 255 for green and zero and this is 0. 0, 255. All right, so this pretty simple and for I believe most of you is you already knew this.09:23So two, white is high intensity for all three channel. It looks white and black is like no intensity at all for all three channels. And we like to represent it using a normalized floating point and it will be one0.09:390 for red and for green will be 0, 1. 0. And then move, move this cursor away. All right, so this. So okay, I'm using my keyboard to move this around, but I realize it distracting. Okay, now it's for blue, 001 white is 111.09:58and Z, Z, z for black. Okay, and now if we mix this RGB channels and can get some kind of other color as well, very. So I just picked three very popular colors. And as a review for yellow, this is mixing of the highest intensity in red and green.10:18So get yellow and S is the green and blue channel mixing and magenta is the, what is it, Uh, is it? It's a red, red and blue, Right? red and blue, almost like kind of purplish. So for this one and then, and if you map that to the float representation, normalized float, by the way, not just any float normalized float normalize between one and zero is one, one, zero, 0, one one and one zero one. So have four pixel, but use 12 values to represent this four pixel image. All right, so for red, another decision have to make is that we know we're going to use three of this pixel to represent a color, but we have to agree upon the order in which RGB are the are arrange.11:45So of course, the most popular common is arranging RGB, but in some other systems, other library, they might makes an assumption of how RGB arranged could be BG, for instance, open CV library that, uh, they assume this this this colors are arranged in a different order. when we actually like to actually display them and if you, uh, so the way that you look at the screen at this moment and it have lots of pixels, but if you you happen to have a microscope with you, you probably don't but you happen to have it Imagine you really really trying to magnify the your screen you're looking at right now.13:27At the end, you're going to see that each pixel actually there, three different lights, LED matric kind of like every pixel is three small lights here, and they will be control separately. And trying to mix the color you actually want.13:44For instance, for this is going to be what, what do I have here? This is going to be red, this is going to be yellow. when I arrange them into a matrix, This is going to be sign, I just kind of write down some words for the remember, okay, black, okay.14:02And then so have four different pixels, and then I kind of put them in this shape. So that look square, even though it has three different lights. So the way that I like, I would draw this, I will be just draw something larger like this.14:16So the, the first pixel is red. So we have four color in this red light. And but for the other one is kind of black. Okay. And then for yellow, we have, what do we have? Yellow is red and green.14:35So we we'll draw red again here, change to red. This is my red light. and green change to green. Do I have green here? This is green, green and blue will be black. So it kind of look like this.14:51And for C is the mixing between green and blue. I mean green and blue. let's TR green. So green is fly bright, and blue is let's use switch one blue, this is blue. Okay. And then for red is black. This lecture introduces image representation in computer vision. It covers various image types (array, matrix, binary, byte array, double array, RGB), explaining their structure and conversion methods, particularly focusing on how to represent them in Excel spreadsheets. The lecture uses small examples to illustrate concepts, then scales up to larger examples in Excel, showing how formulas can mimic code for image manipulation and display, culminating in a demonstration of creating and displaying an RGB image.