← Course overviewFree lesson 1 of 3 · lesson 1 of 22Next lesson →

Unlock all 22 lessons of Introduction to Medical Image ProcessingFull course ₹40,000 · paid

Price & syllabus
Free lesson · 27 min · no payment

What Is a Digital Image: Image Formation and Imaging Modalities

A free lesson from Skill-Lync's Introduction to Medical Image Processing course, on Grow by Skill-Lync · Last updated 16 Nov 2026 · Session: “Module 1 Session 1”

What will you learn in this free Introduction to Medical Image Processing lesson?

Outlines the course on medical image processing, then defines an image as a function of coordinates with values at each location. Explains digital image processing, and classifies imaging by physical principle, including photography, ultrasound reflection, MRI and X-ray attenuation.

  • Course roadmap from sampling to DICOM and 3D Slicer
  • An image as a function of X and Y coordinates
  • What digital image processing means with image input and output
  • How ultrasound, MRI and X-ray form images

What comes after lesson 1?

The next free lesson is lesson 2, “Image Processing Pipeline Stages and the Illumination-Reflectance Model” (34 min). Later in Skill-Lync's Introduction to Medical Image Processing course (paid): lesson 4, Module 1 Session 4; lesson 5, Module 2 Session 1; lesson 6, Module 2 Session 2. The full course has 22 lessons (12 h 42 min of video); see the course overview or its price and syllabus.

Lesson transcript

Machine transcript of the lesson audio (lightly cleaned). Technical terms may contain errors; the video is the reference.

Hello and welcome everyone to this course on medical image processing. In this course, we would be looking into image processing fundamentals, looking into what is a digital image, classification of imaging systems. We also will look into what are the steps while doing image processing. What are the components when you take an image processing system? A system is all that which comes along to process an image and then we look at image formation or image sensing typically for 2D image and also for a 3D image. For medical images, we generally look into how 3D images are being formed and hence some aspect on 3D image formation also would be seen.

The aspects which we follow will be on image acquisition. Once you acquire an image, how do you store it in a computer, right? There are two aspects which goes behind converting an image into a digital image and we look into sampling and quantization. What are the, how do you gauge the quality of an image and then what impairs quality of an image like noise and since this course is oriented towards medical imaging, we then look into a standard for medical image storage and communication called DICOM. We go in some detail about how DICOM file format looks like and then we, Slicer is a open source software which can be used for medical image processing and it is being widely used by the research community and we would see how to read and visualize DICOM files.

So what are the learning objectives? What would be the takeaway from the first week of this course? We would be introduced to what is exactly an image, what is a digital image and once we understand what is an image, how do we go about processing that image and then if there is a system which is used for image processing, what are the various aspects of such an image processing system and the specific aspect about how to acquire an image, how to digitize in terms of sampling and quantization, how to gauge the quality of an image, what metrics can be computed there, what impairs an image quality like noise will also be seen and we come across the medical image format for storage and communication which is the DICOM and then we get into 3D Slicer software to read and visualize DICOM images.

So let's get started with the fundamentals of image processing and to begin with we need to see what exactly is an image. Fundamentally an image is a high dimensional function. We can call it an N dimensional function. To make it easy, we look at it as a three dimensional function to start with. It can have higher dimensions too. But to start with, let's visualize a 3D function where you have X, Y and values at each of these locations. So X and Y are the coordinates on a plane. X is the X axis, Y is the Y axis.

So you have a bunch of X coordinates running in the image, Y coordinates. So it's a grid. And then at each of these XY locations, you have a value for that function and that value of that function is called amplitude. How bright is that image at that location or what is the intensity value of that pixel or that location, right? So that is then the amplitude. So basically, the interesting aspect is the amplitude, but to utilize it in a image format or in a matrix format, you need the grid and that grid are formed by that X and Y coordinates.

So this then comes to the definition of what a digital image is. A digital image is nothing but a finite number of elements. It can be a simple case, a 4 cross 4 or a 32 cross 32 grid of values. At each location of that grid, we have a value which is called a pixel or image element. One can visualize it as a high dimensional function on a grid and that is exactly as a digital image. So how do you visualize this and what makes it easy to perceive an image or how it how does computer perceive an image?

An analog image is how we perceive humans perceive a scene in the natural world. And it is probably having continuous values. And that's why it's called analog. Now, from for a computer to process this, it needs to convert it into a quantized representation, quantized in which all locations you want to have reading of that image taken. And what is that representation in each location in a quantified manner? You cannot have a continuous value there and the continuous value is an analog format. So you need some way to quantify what is that pixel value at that location.

So the first is the sampling part, which decides how coarse or fine is the X, Y grid. So you have a bunch of different X coordinates and Y coordinates here. Here we see about 10 different X coordinates and maybe 10 different Y coordinates. One could have 100 here. One could have 1000 here. So then if you are having 100, you are doing a fine sampling. If you are having 10, you are doing a coarse sampling. More about sampling, we will anyway see when we get to the section on sampling, image sampling.

Now, once we decided that this is the locations where we need to read the image out, we go to the next step of quantizing the amplitude values. We saw the definition in the previous slide that an image is nothing but a multidimension function with X, Y. So here is the X and the Y. And at each of the location, we have an amplitude. But the amplitude can be a continuous in the analog scenario. But to represent it in the computer, we then need to then say, what is, how do we quantify that?

And then that is called pixel quantizations. So one can define, we take this many buckets within which we quantify. We can say 0 to 10, 10 to 20, 20 to 30 are all certain buckets. And then if a value comes 15, we then map it in the 10 to 20 bucket. So it is about how do you, how many distinct values do you want to represent? In the most naive way, you can represent an image as a binary. You only say if it is a 1 or a 0.

And if you go to a grayscale image, you have 256 different values because they use 8-bit image. More of it we will learn later. But quantization is sampling in your amplitude domain. Sampling is how do you grid structure the XY domain. So this is the XY domain. How do you digitize it? And this is the amplitude domain. How do you digitize it? I hope with this definition, you get a mental picture of what an image is. End of the day, what an image is a grid of numbers.

May a computer sees an image as a matrix. It has a certain number of rows, certain number of columns. And each of these rows, columns have values. But that is a very, very crude representation of an image. And image processing experts would completely not agree that an image is just a matrix of numbers. Image has much more than a matrix of numbers. And we would also see, especially in the context of medical image, why such a representation of just a matrix doesn't suffice to represent a medical image.

There is far more information that is needed. And this we would see also in some level of detail when we get into the DICOM and how DICOM images are represented. So what is digital image processing? Now that we understand what an image is, what are we talking about image processing? And what is digital image processing? Now, digital image processing is when your input is an image. You take an image with your mobile phone camera. You can take an image with, say, ultrasound system in a hospital or an X-ray.

You go to a hospital, they take an X-ray. So the input is an image. You have a medical image in front of you. And then you want to process that image and get some insights into it. And the output could be another image. It can be an enhanced version of the image because you want to see certain structures which are present in the image. One could also go a step further and say we want to not just have an image. We want to extract attributes in the image.

Attributes could be like how many characters are there in this slide. That's an attribute of this image. How many dogs are there in a scene? That's an attribute. So it also involves processing which includes object recognition. You need to find out if you say how many dogs are present in an image, you need to know what does a dog look like. And then maybe put a bounding box around a dog and then count all the number of dogs. Then you come to the aspect of the attribute of the image saying how many dogs are present.

So there are multiple layers of complexity. And we'll see that also in the subsequent slide that there is low-level image processing. There is mid-level image processing and advanced-level image processing or high-level image processing. And what each of these levels try to do. And this is more a crude classification to say what are the challenges in low, mid, and high level of image processing. Now why is there a digital aspect to it? The digital because the images are being processed by a computer. And hence it is called digital image processing.

And there is a very strong relationship between digital signal processing and digital image processing. DSP as it is generally known is about how do you process a signal with signal processing techniques. And the extension of a signal. I don't know. To two-dimension is an image. And hence, many of the concepts of digital signal processing also applies to images in a maybe slightly different context. But the fundamental premise still remains the same. So now that we know what is digital image processing and we understand what is going to be the input, the input is going to be an image.

What is going to be the output? The output could be an image. The output could be an attribute or a description. The input could also be a description and the output from the description could be an image, right? So here we have a two-cross-two grid and it makes some amount of time to go in level of detail to understand what exactly are we looking at the various adjacencies of image processing. As I mentioned, if the input is an image and the output is also an image, then it's predominantly digital image processing.

And so here you can take an image like a photograph of maybe the Alps, the Swiss Alps, and then you enhance the image to make the snow look better. Or you see, take a picture of an ocean and then you want to make the waves look better. So the input is an image of an ocean, the output is waves behaving well represented. So that is the digital image processing part. Now we come to the second part of it where the input is an image and the output is a description.

So we have an input of, say, a scene like a forest and the output is like how many trees are there. Another example, it can be an image, a photograph of students sitting in a classroom and in the output you want to then say how many students are sitting there, a count of the number of students. So here the output is a description and not necessarily an image, a metric, right, that you compute on the image. How many, what is there in the image, right? These are descriptions of the image.

One of the very good examples of this is the attendance, automatic attendance mechanism, where they explore finding out how many students have come to a class, has a particular student attended the class or not. These come under computer vision domain, right? The input is an image and the output is a description. Now we come to the inverse problem of this. The input is a description and the output is an image. It is slightly more different than image processing and computer vision. Here you then say I want a vase which is maybe three meters tall and having about two meters width and maybe elliptical in shape, right?

You give such descriptions and then those descriptions are then either manually done using a software to create that object and or it can be just fed into a system which understands your description and you can create that object and that is computer graphics. So, graphics is starting from a description saying I want two volleyball players playing volleyball in the beach. It's a description. Now, then you have to go about designing that scene with a beach, sand, a volleyball net, maybe a volleyball and then a couple of players on either side.

That then is the scene that you create based on the description provided and computer graphics has immense number of tools to create such an image. The final aspect is your input is a description and output is a description. One could call it very crudely as artificial intelligence. Artificial intelligence is much broader than this. But one could in some fair way understand this as the input is a description saying yes, here is a scene with so many vehicles going in the street and the output is like is it how far am I from a particular object in an automatic driving scenario?

So, AI is actually much broader than this. It is to learn more about AI when you cover machine learning in detail. But for now, this can be a placeholder to then say yes, this could be one way to look at artificial intelligence. So, I hope the four quadrants are very clear what is image processing, what is computer vision, what is graphics and what is AI. As I mentioned in the previous slide, there are multiple levels of image processing. You start with a low level which is easiest to understand and as I mentioned here, the input and output are indeed images.

You have an image, you have an image, you want to remove the noise in the image because noise has arised due to whatever acquisition mistakes. The other is you have an image which is very dull, you want to make it brighter, you want to make it more sharper. So, those are low level image processing where the input and output are images. In our previous slide, we defined this as image processing, but that's really the low level image processing. Now, we come to the mid-level image processing where the images are input and the outputs are attributes and then I was giving the example of counting the number of students in a classroom and seeing is a particular person attended the class or not.

And then the high level processing is making an ensemble of recognized objects, right? So, you have a self-driving car, right? And then the self-driving car has to understand an ensemble of various things which are there in the scene. There could be trees, there could be pedestrians walking, there could be cyclists, there could be any other object on the street which has to be understood. And then you have to make decisions or cognitive functions on top of that, right? So, the example given is autonomous navigation. You want to navigate a car with no driver in a scene which is dynamically changing.

A pedestrian may walk at any point in time. A motorist, two-wheeler can come any which way from any corner. And then these has to be understood and recognized. And so, it is making sense of an ensemble of objects. There are multiple objects. And making to able to make cognitive decisions based on that. Cognitive meaning, can I drive now continuously? Should I pause? Should I brake? How quickly can I go? How slowly should I go, right? These are cognitive functions that can be taken based on. So, those are the three levels.

And it can be slightly confusing to say that high level is, is it indeed image processing or does it come under computer vision? The boundaries are a little fuzzy. There is no stop where they say, this is it, image processing. And from here on, it is computer vision. Advanced high level image processing is same as computer vision, right? There are a lot of overlaps there. And then the advanced computer vision will pick off from there and then make it even more complicated in making higher level attributes of the scene to be computed, right?

So, now we come to what are the different imaging modalities, right? So, when I say what are the imaging modalities, what are the sources from which we can capture images? Source of the electromagnetic spectrum, right? Light falls on an object. The object reflects the light. And then you capture an image. But what is the intensity, frequency of that light? What is the wavelength of the light? Based on that, the imaging modalities are specified or are different. So, if you see on the bottom graph here, you have varied ranges of frequencies.

And this could also be represented as varied wavelengths. And each of these different frequency levels have a particular imaging modality. So, if you go to the very high frequency, they are called gamma rays. Slightly lesser frequency are the X-rays and ultraviolet rays. And then much lower frequency are the microwaves and radio waves, right? So, in the subsequent slide, we will see examples of each of these imaging modalities. I mean, even from this slide, we see the different imaging modalities. So, if you have very high frequency that like gamma rays that is used in and we are giving examples of medical imaging in this case that is used in PET.

That is expansion for PET is positron emission tomography. And a way to abstract both PET and SPECT, single photon emission for tomography and positron emission tomography is they call in medical domain as nuclear medicine, right? One can also use gamma ray imaging in astronomy. Now, if you come to slightly lesser frequency, then you get X-rays, which many of you would have observed when you go for getting a scan done of your of a fall or of the chest to make an analysis. If there is any inflammation in the chest, generally the doctors advise an X-ray to be taken.

And here the frequency is not that high. Frequency also means that they are ionizing. Ionizing means they kind of change the characteristics of the object being imaged. So, hence it is recommended that in medical domain, you shouldn't get exposed to very high ionizing radiation like gamma rays and X-rays unnecessarily. If there is a necessity for diagnosis, you can go get an X-ray done. But you shouldn't do an X-ray. I feel like getting an X-ray done. That is never done that way, right? X-ray does cause ionizing radiation, which is some amount of damage to that tissue which is exposed to X-ray happens because of the high energy it has.

Whereas if you come to, and I mean the subsequent one is ultraviolet imaging and this is used in microscopy where you have maybe an image of a blood smear put on a glass slide and you want to observe each cell. And here they use ultraviolet imaging. That is also not very, very high in frequency but higher as compared to our physical visible range. So, if you see the graph below, the visible range of images is a very small portion as compared to the entire range in which one can acquire images.

We generally think images are what we can perceive by our eyes. But if you really see there is a lot of images which can be. Quad, using frequencies, frequencies much higher than the visible light, which are here, gamma rays, x-rays, ultraviolet, or much lower than the visible light, which are the infrared, microwaves, and radio waves. So some more details about these imaging modalities. As I mentioned, the visual and infrared are what images we generally capture using a mobile phone or a camera, a digital camera or a camcorder.

Those are the visible images. But then if you really see the range of frequencies or range of wavelengths where we capture visible image, it's a very, very narrow band. Going beyond the visible range is the microwave range and the radio band. And we also have radio waves, which one of the modality there is the magnetic resonance imaging, where you use a magnet to find out how many hydrogen particles are there in a given object. You can call it abstractly as how much water content is present. But in reality, what is being imaged is how much hydrogen is present in given object, which is being imaged.

And then there is also the other modality, which is the electron microscopy. Again, this operates in a much, much lower frequency per se. So there are two different kinds of imaging. Again, this comes more relevant when we talk about digital image processing in the medical context, right? So medical images are not taken by taking a point and shoot camera, looking at a scene and click a button, throw a flash and capture an image. I don't think we have reached that stage or we would not reach that stage even in the near future where such a thing can be done.

So a direct imaging is when you have an image which is when acquired, already acquired in a recognizable format. So you look at an image and then you know that this is what I captured, right? You would throw a flash, click your digital camera like your mobile camera, and then you get an image on your screen. And then you can immediately see what you have acquired. And this can be done in parallel or serial mode. Parallel is how our human vision system works, right? So you look around, you kind of see multiple objects in one go.

It's not that you see object one, then you have to move and then see object two. How much ever the field of view, the eye can cover, all the objects are simultaneously acquired, processed, understood in one go, right? Serial acquisitions are more we go step by step, line by line, row by row, column by column to then get an image done. And there are applications like micro densitometer, which the application for this is to find out what is the density of bones in, say, for example, on the feet to find out disease scenarios like osteoporosis.

So here you go pixel by pixel, maybe scanning one row at a time, and hence it is more a serial acquisition. And then the indirect imaging, what is not direct, right? So what is direct is when you click a button, you throw light and you capture the image, you get that. But in medical imaging, generally there is a image formation process. So you kind of rely on the physics of a particular system, imaging system to come about how to get that image reconstructed and the image directly cannot be visualized.

So if you go into a CT room, like a radiology place in a hospital and they recommend you to take a CT, you can't visualize the CT scanner's data and figure out what that is. Because there is a process required for reconstructing, recon, as they call in medical parlance. There is a process of reconstructing the image and the reconstruction can be based on some transforms like radon transform or Fourier transform. MRI, they use the Fourier CT, they use the radon transform. And only after you do that transform, can you even recognize what is it that you have acquired, right?

And that goes, the reason for such as acquisition step, the reason is that there is a physics behind this whole imaging process. What is the source? What is the detector? How is the image being acquired? And that entails that there is a recon step generally required to view the image. And those kinds of reconstruction requiring protocols come under indirect imaging. We briefly saw about various imaging modalities. Here is a slightly more detailed representation of various kinds of images based on, this is a classification of images based on physical principles, right?

So I wouldn't go through all the various aspects, but I just look up and give some details about few of them, right? So let's take the simple visible one, right? You can then say we saw about wavelength and we saw about frequency and we saw various imaging modalities. But if you have to look at visible range, what are the things that you can do, right? That comes under general optical reflectance, right? So light falls on an object, the object reflects some light back, you capture that and digitize that, and that gives you an insight about what that image is, right?

And this comes from this mechanism of processing an image from raw data and making it look like an image comes under the general scenario of optical reflectance, right? The rest of it, you have acoustic reflectance, you pass sound, you get sound back, and then that comes under ultrasound. Then you have attenuation, you pass a high intensity, high frequency radiation, electromagnetic radiation through an object. When you pass x-ray, you have an object, and then there is some structures like bone which impair the x-rays to pass through, and the other locations allow the x-ray to pass.

And that then forms a map of attenuation. Where are objects, the x-rays being obstructed? Where are x-rays allowed to pass freely, right? And with that information, one can create an image, and that is how an x-ray is being formed, right? So I have just covered a couple of these scenarios. One of them is general photography context, a little bit on medical ultrasound based on sound and its reflectance, and so some bit based on the electric and magnetic properties like how MRI gets computed. So these are the various categories for classification based on the physical principles.

How does the energy source, right, the electromagnetic spectrum source come from? Is it coming from sound? Is it coming from impedance or blocking some energy sources? Or is it due to magnetic properties of certain objects, electric properties of certain objects, and so on, right? So there's varied ways of getting an image acquired. So this finishes our session one for module one.

Want the rest of Introduction to Medical Image Processing?

Get the full course syllabus and fees.

The full course is paid (₹40,000). We'll send the syllabus and call you back within one working day.

Talk to a career counsellor

Free 20-minute call. We'll map the right course to where you want to work.

By submitting you agree to our Terms and Privacy Policy.

Full course · paid

Unlock the full Introduction to Medical Image Processing course

The remaining 19 lessons, mentor-reviewed projects and a certificate come with the paid course (₹40,000). Leave your details: we'll send the syllabus and fees and call you back.

Get the syllabus and fees

A counsellor calls you back within one working day.

By submitting you agree to our Terms and Privacy Policy.

Grow by Skill-Lync · free lesson

Start watching

Free, no payment. Leave your details once and every free lesson on Grow plays straight away. A Skill-Lync counsellor may call you about the full Introduction to Medical Image Processing course.

By submitting you agree to our Terms and Privacy Policy.

Talk to a career counsellor

Talk to a career counsellor

Free 20-minute call. We'll map the right course to where you want to work.

By submitting you agree to our Terms and Privacy Policy.