introduction to digital image processing using matlab

Raymond Phan – Ph.D. Candidate

Department of Electrical and Computer Engineering Distributed Multimedia Computing Research Lab

EPH 408 [email protected]

Some slides were taken from Prof. R. A. Peters PPT slides: http://www.archive.org/details/Lectures_on_Image_Processing

mailto:[email protected]

http://www.archive.org/details/Lectures_on_Image_Processing

Topics Covered in this Presentation 1st Hour: 6:10 p.m. – 7:00 p.m.

Introduction to Digital Images and in MATLAB Quick intro to myself What is MATLAB? Where to get MATLAB and how to run Basic I/O Reading and writing images Accessing pixels and groups of pixels Resizing Images Rotating Images …break for 10 minutes!

Topics Covered in this Presentation 2nd Hour: 7:10 p.m. – 8:00 p.m.

Operations on Digital Images Simple contrast and brightness enhancing Intro to image histograms Advanced enhancing using image histograms Intro to convolution in images Blurring / Smoothing images Edge detection Sharpening images …break for 10 minutes!

Topics Covered in this Presentation 3rd Hour: 8:10 p.m. – 9:00 p.m.

Applications of Image Processing Segmenting simple objects Noise Filtering Simple image stitching using template matching

NOTE! For a more comprehensive MATLAB tutorial, check:

http://www.ee.ryerson.ca/~rphan/ele532/MATLABTutorial.ppt

You can access the slides, images and code at: http://www.rnet.ryerson.ca/~rphan/IEEEDIPTalk


http://www.rnet.ryerson.ca/~rphan/IEEEDIPTalk


Introduction to Digital Images and in MATLAB Quick intro to myself What is MATLAB and where do I get it / find it? Intro to Digital Images Basic I/O Reading and writing images Accessing pixels and groups of pixels Resizing Images Rotating Images …break for 10 minutes!

Introduction of Myself Started in 2002 in the B.Eng. – Computer Engineering

program Program only started at this time Relatively new Graduated in 2006

Started my M.A.Sc. in 2006 – ELCE Finished in 2008 Winner – Ryerson Gold Medal (2008 – SGS)

Started my Ph.D. in 2008 – ELCE Will finish before my back gives out Currently a 4th year Ph.D. Candidate 2010 NSERC Vanier Canada Graduate Scholar

Introduction of Myself (2) Research Interests

Digital Image Processing, Signal Processing, Multimedia, Computer Vision, Stereo Vision, 3DTV, etc.

M.A.Sc. Thesis – Content-Based Image Retrieval System Featured in the Toronto Star (January 4th, 2008) Searching for images using actual images, rather than

keywords Current Ph.D. Thesis – Faster and more accurate 2D to

3D conversion

What is MATLAB? MATLAB Stands for MATrix LABoratory Created by Cleve Moler @ Stanford U. in 1970

Why? Makes linear algebra, numerical analysis and optimization a lot easier

MATLAB is a dynamically typed language Means that you do not have to declare any variables All you need to do is initialize them and they are created

MATLAB treats all variables as matrices Scalar – 1 x 1 matrix. Vector – 1 x N or N x 1 matrix Why? Makes calculations a lot faster (will see later)

What is MATLAB? (2) How can I get and/or use MATLAB? 3 ways:

1) Find it on any of the ELCE departmental computers Log in to your account, go to Applications Math

MATLAB R2010b 2) Use any Ryerson computer on to the ACS network

Log in with your Matrix ID and Password, then go to Start MATLAB R2010b

3) Install it on your own laptop Go to http://www.ee.ryerson.ca/matlab for more details You must be on the Ryerson network to sign up for an account After, you can download MATLAB from anywhere

MATLAB uses the Image Processing Toolbox (IPT) Should already be installed with MATLAB!

http://www.ee.ryerson.ca/matlab

Intro to Digital Images Digital Image

a grid of squares, each of which contains a single color

each square is called a pixel (or picture element)

Color images have 3 values per pixel; monochrome / grayscale images = 1 value/pixel.

Intro to Digital Images – (2) A digital image, I, is a mapping from a 2D grid of

uniformly spaced discrete points, {p = (r,c)}, into a set of positive integer values, {I( p)}, or a set of vector values, e.g., {[R G B]T(p)}.

Each column location of each row in I has a value The pair (p, I(p)) is a “pixel” (for picture element) p = (r,c) pixel location indexed by row r & column c I(p) = I(r,c) Value of the pixel at location p If I(p) is a single number I is monochrome (B&W) If I(p) is a 3 element vector I is a colour (RGB) image

Pixels

Intro to Digital Images – (3) Monochromatic Case:

We call the values at each pixel intensities Smaller intensities denote a darker pixel Bigger intensities denote a lighter pixel

Colour Case: Think of a colour image as a 3D matrix First layer is

red, second layer is green, third layer is blue Why RGB? Trichromacy theory All colours found in

nature can naturally be decomposed into Red, Green and Blue This is basically how CCD cameras work!

The three element vector tells you how much red, green and blue the pixel is compromised of (i.e. [R G B]T = [0 255 0] No red, no blue, all green

Intro to Digital Images – (4)

=

=

614312

bluegreenred

)( pI

( )( )( )277,272

col,row,

===

##crp

Pixel : [ p, I(p)]

Pixel Location: p = (r , c) Pixel Value: I(p) = I(r , c)

Intro to Digital Images – (5) How do digital cameras take images (very basic)?

Uses sampling and quantization What we see now through our eyes is continuous

There is essentially an infinite amount of points that comprise our field of view (FoV)

Not good, because we want to store this information! We first need to sample the FoV Transfer the FoV to

a rectangular grid, and grab the colour in each location of the grid

Intro to Digital Images – (5) Sampling and Quantization

sampled real image quantized sampled & quantized

pixel grid column index

row

inde

x

Intro to Digital Images – (6) We’re not done yet! There are also an infinite number

of possible colours We will now need to quantize the colours Quantizing will reduce the total number of colours to a

smaller amount Key Quantize accurately so that we can’t tell much

difference between the original image and the quantized one

A digital image is essentially taking our FoV and performing a sampling and quantization Values are now discrete and positive

Intro to Digital Images – (6) Sampling and Quantization

sampled real image quantized sampled & quantized

pixel grid column index

row

inde

x

Intro to Digital Images – (7) Digital images store their intensities / colour values as

discrete and positive values Usually, digital images need 8 bits for B & W and 24-

bits for colour (8 bits for each primary colour) B & W – 0 for Black and 255 for White All integers Colour – 0 to 255 for Red, Green and Blue All integers Note: We can consider a colour image as three 2D images

Without compression, files would be very large! Compression algorithms (PNG, JPEG, etc.) eliminate

extra information to reduce the size of the image

R/W Images in MATLAB So we have an image file… how do I access the info? Open up MATLAB and change working directory to

where image is stored Use the imread() function

im = imread(‘name_of_image.ext’)

Use single quotes, and type in the full name of the image with its extension (.bmp, .jpg, etc.)

im will contain a 2D matrix (rows x cols) of B&W values or a 3D matrix (rows x cols x 3) of colour values

Matrix corresponds to each pixel in the digital image for B & W, or a colour component of a pixel in colour

R/W Images in MATLAB – (2) How do I access a pixel in MATLAB B&W case?

pix = im(row,col); row & col: Row & column of the pixel to access pix contains the intensity value Access elements in an array by round braces, not square!

For you C buffs Indexing starts at 1, not 0! How do I access a pixel in MATLAB Colour case?

pix = im(row,col,1); Red colour value pix = im(row,col,2); Green colour value pix = im(row,col,3); Blue colour value 3rd argument 3rd dimension of matrix Only grabs one colour value at a time!

R/W Images in MATLAB – (3) How can I get the RGB pixel entirely? Use the :

command pix = im(row,col,:);

: means to grab all values of one dimension However, this will give you a 1 x 1 x 3 matrix… we just

want an array! Call the squeeze() command pix = squeeze(im(row,col,:));

Now a 3 x 1 vector. To access R, G and B values, do: red = pix(1) Red, gr = pix(2) Green, blue = pix(3) Blue

R/W Images in MATLAB – (4) So I know how to get pixels; how can I modify them in

the image? Easy! Just go backwards For a B & W Image do: im(row,col) = pix;

For a colour image, do either: im(row,col,1) = red; im(row,col,2) = green; im(row,col,3) = blue; or im(row,col,:) = [red; green; blue] or im(row,col,:) = rgb; %rgb - 3 x 1 vector

R/W Images in MATLAB – (5) How do I access a subset of the image?

How do I grab a portion of the image and store it into another variable?

Do the following for monochromatic images: im2 = im(row1:row2,col1:col2);

Do the following for colour images: im2 = im(row1:row2,col1:col2,:);

This will grab a rectangular region between rows 1 and 2, and columns 1 and 2

e.g., if I wanted to get rows 17 – 31, and columns 32 – 45 for colour, do: im2 = im(17:31,32:45,:);

R/W Images in MATLAB – (6) So I know how to get pixels; how can I display images? Use the imshow() command

imshow(im); im: Image loaded into MATLAB Shows a new window with the image in it

If you do: imshow(im,[]) For monochromatic: Smallest intensity becomes 0 and

largest intensity becomes 255 for display For colour: Apply the above for each colour channel

Every time you use imshow, the image you want to display is put in the same window… so what do you do?

R/W Images in MATLAB – (7) Use figure command to create a new blank window

Then, run the imshow command to display the image on the other window

We can also do: imshow(im(:,:,1)); Shows red channel imshow(im(:,:,2)); Shows green channel imshow(im(:,:,3)); Shows blue channel

(:,:,1) means grab all of the rows and columns for the first layer (i.e. red), etc.

R/W Images in MATLAB – (8) When showing the red channel:

Darker pixels mean there isn’t much red in that pixel Lighter pixels mean there is a lot of red in that pixel

Same applies for green and blue! How do I save images to disk? Use imwrite()

imwrite(im, ‘name_of_image.ext’, ‘EXT’);

im image to write to disk name_of_image.ext Name of the image ‘EXT’ Extension of the file (‘JPG’, ‘BMP’, ‘PNG’, etc.)

Demo Time #1! Reading in Images

Accessing Pixels Changing Image Pixels

Obtaining a Subset Display Images Monochromatic and Colour

Displaying Colour Channels Separately Writing Images to File

Resizing Images One common thing that many people do is resize

images i.e. Make an image bigger from a smaller image, or make

an image smaller from a larger image How do we resize images in MATLAB? Use the imresize command How do we use it?

out = imresize(im, scale, ‘method’); or out = imresize(im, [r c], ‘method’);

For both methods im is the image we want to resize, and out is the resized image

Resizing Images – (2) Let’s look at the first method: out = imresize(im, scale, ‘method’);

scale takes each of the dimensions of the image (# of rows and columns), and multiplies by this much to determine the output e.g. If we have an image that is 20 rows x 40 columns: If scale = 2, the output 40 rows x 80 columns If scale = 0.5, the output 10 rows x 20 columns

method determines the type of interpolation when resizing Important when making an image bigger

Resizing Images – (3) When we are making an image bigger, we are trying to

create an image with a lack of information present There are three main types of interpolation

Nearest Neighbour method = ‘nearest’ Uses the best pixels that are near the original pixels and fills in

missing information Bilinear Interpolation method = ‘bilinear’

Uses linear interpolation in 2D to fill in missing information Bicubic Interpolation method = ‘bicubic’

Uses cubic interpolation in 2D to fill in missing information

Usually, bicubic is known to have the best accuracy

Resizing Images – (4) Now, let’s take a look at the second method for resizing out = imresize(im, [r c], ‘method’); This routine will resize the image to any desired

dimensions you want You can customize how many rows and columns the

final image will have Example: To resize a 130 rows x 180 columns image to

65 rows x 90 columns, with bilinear interpolation, do: out = imresize(im, [65 90], ‘bilinear’);

We can also do! out = imresize(im, 0.5, ‘bilinear’);

Rotating Images Suppose we want to rotate an image How? out = imrotate(im, angle, ‘method’);

im: The image we want to rotate angle: How much we want to rotate the image

Angle is in degrees! Positive angle denotes counter-clockwise rotation, and negative angle is clockwise

When rotating, there will inevitably be some missing information method is like before with resizing

out: The rotated image Example: Let’s rotate CCW by 45 degrees by bilinear: out = imrotate(im, 45, ‘bilinear’);

Demo Time #2! Enlarging Images Shrinking Images Rotating Images

Cont. & Brig. Enhancement First real application Brightness enhancement

How do we increase /decrease brightness of an image? 1 way: Just add or subtract a brightness to every pixel How? im2 = im + c; or im2 = im – c; c is any constant (0 – 255) Takes c and add / subtract to every pixel in the image Adding / subtracting makes the image brighter / darker Round off occurs if out of range (i.e. set to 255 if > 255/set to 0 if < 0)

What about another way? We can also scale the image by a constant Do: im2 = c*im;

If c > 1, increasing brightness If c < 1, decreasing brightness

Cont. & Brig. Enhancement (2) So we covered brightness… what about contrast? Contrast How well you can see the objects from the

background When performing brightness enhancing, you’ll notice

that it looks “white washed out” Poor contrast We can do a contrast enhancement to make objects

look better, leaving background relatively unaffected How? Use the power law s = rγ

r is the input pixel intensity / colour and s is the output intensity / colour

Cont. & Brig. Enhancement (3) For each pixel, apply this equation to each of the

intensities / colours For colour images,

apply to each channel separately

Exponent γ determines how dark or light the output is γ > 1 Make darker γ < 1 Make lighter

Cont. & Brig. Enhancement (4) How do we apply the power law in MATLAB?

Use imadjust out = imadjust(im, [], [], gamma);

im: Input image to contrast adjust Ignore 2nd and 3rd parameters Beyond the scope of

our talk gamma: The γ exponent that we’ve seen earlier out: The contrast adjusted image Example use: If γ = 1.4, we do: out = imadjust(im, [], [], 1.4);

Demo Time #3! Contrast and Brightness Enhancement

Intro to Image Histograms We can perform more advanced image enhancement

using histograms Before we cover this… we should probably cover what

histograms are! So, what’s a histogram? It measures the frequency, or how often, something

occurs Let’s look at a grayscale image for now Expressed as H(x) = q, x is an intensity- [0,255] for 8 bits This tells us that we see the intensity value of x for a

total of q times

Intro to Image Histograms – (2) Example: Let’s take a look at an grayscale image

There are ~1500 pixels with a gray level of around 10 There are ~1200 pixels with a gray of around 170, etc.

Intro to Image Histograms – (3) How do we create histograms in MATLAB?

imhist(im); %im is read in by imread

Assuming im is a grayscale image If we want this to work with colour images, we will have

to display the histogram of each colour channel by itself imhist(im(:,:,1)); % histogram for red

imhist(im(:,:,2)); % histogram for green

imhist(im(:,:,3)); % histogram for blue

The (:,:,1) means that we should grab every row and column from the red channel, etc.

Intro to Image Histograms – (4) Grabbing all of the pixels in any channel will produce a

grayscale image, which can be used for imhist We can also call the histogram function by: h = imhist(im);

Will create a 256 element array, where the (i+1)’th element contains how often we see the grayscale i.

Histograms give us a good insight on image contrast If the histogram has too many values towards the left Image is too dark

If the histogram has too many values towards the right Image is too bright

If the histogram has too many values in the middle Image looks very washed out

Advanced Enhancement – (1) Above cases are when the image has bad contrast A good image should have good contrast

i.e. The histogram should have an equal number of pixels over the entire histogram (a flat histogram)

If an image has bad contrast, we can try to correct it by doing histogram equalization Tries to make the image’s histogram as flat as possible

for good contrast Stretches the histogram out For you probability nerds If you divide each

histogram entry by the total number of entries, this forms a Probability Density Function (PDF)

Advanced Enhancement – (2) Histogram equalization seeks to modify the PDF so that

all possible events (pixels) are equally likely to occur How do we perform histogram equalization? out = histeq(im); %im given by imread What about for colour images?

You will have to perform histogram equalization on each channel individually, then merge together

r = im(:,:,1); g = im(:,:,2); b = im(:,:,3); out(:,:,1) = histeq(r); out(:,:,2) = histeq(g); out(:,:,3) = histeq(b);

Demo Time #4! Displaying and Calculating Histograms

Contrast and Brightness Enhancement by Histograms

Intro to Convolution Before we proceed, we need to understand what

convolution is for a digital image Probably learned in ELE/BME 532, ELE 792, etc…. Bleck! But! Actually very simple when dealing with 2D images

Preliminaries: First, we need to create an N x N matrix called a mask The numbers inside the mask will help us control the

kind of operation we’re doing Different #s allow us to blur, sharpen, find edges, etc. We need to master convolution first, and the rest is easy!

Intro to Convolution – (2) Steps:

1) For each pixel (r,c) in our image, extract a N x N subset of pixels, where the centre is at (r,c)

Example: 9 x 9 subsets shown below @ various locations

Intro to Convolution – (3) Another example: Let’s grab a 3 x 3 subset, located at

(r,c) = (6,4) Pixels are a, b,

… g, and centre is e

2) Take each pixel in the subset, and do a point-by-point multiplication with the corresponding location in the mask

Intro to Convolution – (4) Example: Our 3 x 3 subset has pixels:

Our mask has the following quantities

=

ihgfedcba

G

=

rstuvwxyz

H

Intro to Convolution – (5) When we do a point by point multiplication, we will

now have 9 numbers: a*z, b*y, c*x, d*w, e*v, f*u, g*t, h*s, i*r

3) Create an output image and: a) Add up all of these values b) Store the output at (r,c) (i.e. the same row and

column location as the input image) in the output image Example:

out = a*z + b*y + c*x + d*w + e*v + f*u + g*t + h*s + i*r, then store out into (r,c) of the output image

Intro to Convolution – (6) Essentially, convolution is a weighted sum of pixels

from the input image within the subset Weighted by the numbers in your mask

Each pixel in the output image is this weighted sum Do this for each location of the input image and assign

to same location in the output image What about for colour!?

Do convolution on each channel separately Note: If the output value is floating point (decimal),

we must round in order to make this an 8-bit number Here’s one more example to be sure you understand…

Intro to Convolution – (7)

Blur / Smooth Images So now what? Let’s try blurring/smoothing images Blurring? Think of a camera that is out of focus Why should we blur? Minimize sensor noise Noise High Frequency

When we blur, we are essentially low-pass filtering, eliminating high frequency content

High frequency content Edges By getting rid of the edges, we blur the image

How do we blur an image? Try an averaging filter

Blur / Smooth Images – (2) Averaging? 1st, create an output image to store results

1) For each pixel (r,c) in the image, look at a N x N neighbourhood / subset of pixels that surround (r,c)

2) Add up all pixel values in the neighbourhood and divide by N2 (number of pixels in neighbourhood)

3) Take this new value and store it into same (r,c) location

Notes: This works for B & W images What about for colour images?

Blur / Smooth Images – (3) Remember, colour image can be seen as a 3D matrix 3D matrix 2D matrices with layers

1st layer Red values 2nd layer Green values 3rd layer Blue values

So, we can blur each layer independently, and use the RGB values after the blurring of each colour layer

The bigger the neighbourhood, the more the blurring How can we efficiently implement this?

Blur / Smooth Images – (4) Create a mask for convolution to efficiently do this Mask: Same size as desired subset / neighbourhood Mask will contain numbers used to generate our result Examples:

3 x 3 Averaging 5 x 5 Averaging Mask Mask 9 x 9 Averaging Mask

111111111

91

1111111111111111111111111

251

111111111111111111111111111111111111111111111111111111111111111111111111111111111

811

Blur / Smooth Images – (5) These masks make sense. Why?

Let’s look at the 3 x 3 case: Example: Suppose our 3 x 3 neighbourhood is:

Our 3 x 3 averaging mask is:

=

987654321

G

=

=

91

91

91

91

91

91

91

91

91

111111111

91H

Blur / Smooth Images – (6) The output should be:

out = (1)(1/9) + (2)(1/9) + (3)(1/9) + (4)(1/9) + (5)(1/9) + (6)(1/9) + (7)(1/9) + (8)(1/9) + (9)(1/9) out = (1/9)(1 + 2 + 3 + 4 + 5 + 6 + 7 + 8 + 9) out = (1 + 2 + 3 + 4 + 5 + 6 + 7 + 8 + 9) / 9

… isn’t this exactly the same thing as averaging!? You are essentially adding up all of the pixels in the

neighbourhood, and dividing them by the total number of pixels (9)

Note: If we get a decimal number, we round to be sure that we have an integer number

Blur / Smooth Images – (7) Understanding was the hard part! In MATLAB, this is very easy! First, we need to create the mask

We do this by calling fspecial() mask = fspecial(‘average’, N);

mask contains the averaging mask to use First parameter specifies we want an averaging mask N specifies the size of the mask Bigger the N, more

blurring we get

Blur / Smooth Images – (8) Next, call a command that will perform this multiply

and sum command for each pixel in the image Use the imfilter() command out = imfilter(im, mask);

out: output image (this case averaged output) im contains the image we want to blur mask: Convolution mask to use (this case averaging)

Nice little note imfilter() works on both grayscale and colour images For colour, this automatically performs convolution on

each channel individually and combines after

Blur / Smooth Images – (9) imfilter() performs image filtering using masks

Essentially convolution Filtering Produce an output image that changes the

frequency content of the image Blurring, Sharpening, Detecting Edges, etc.

Why are masks important? You change the size of the mask, and values in the mask

and you get different results! Essentially how most filters in Adobe Photoshop and

GIMP work

Demo Time #5! Blurring / Smoothing Images

Edge Detection Next? Edge detection… what is an edge? Edges are sudden discontinuities appearing in images

A.K.A. sudden jumps in intensity or colour in images How can we detect edges? Very simple: Think Calc.

Find the derivative of each point in the image When slope is very high, that means you found an edge

But, derivative is only for 1D signals… what about 2D!? You must take the gradient Derivative in both the x

and y directions You combine both of these directions to create your

final derivative function

Edge Detection – (2) How do I take the derivative? We perform convolution, but

with a different mask Corresponds to discrete approximation of the derivative

Two possible masks we can use: Prewitt 3 x3 Mask Sobel 3 x 3 Mask

These masks detect changes in the horizontal direction If there are no changes, then weighted sum of the pixels in 3

x 3 neighbourhood should have same value Gradient = 0

−−− 111000111

−−− 121000121

Edge Detection – (3) Prewitt Mask Normal derivative Sobel Mask Puts more weight on the central pixels How does this detect horizontal changes?

When using the mask, if there is a change, there is a huge change above the zero line, and below the zero line

So when performing correlation, it can detect horizontal changes

How do we detect vertical changes? Take the transpose of each mask! Tranpose Interchange rows & columns of the mask

Edge Detection – (4) How do we combine the horizontal and vertical

gradient information? Remember from Vector Calculus:

So, we take the horizontal gradient image, and square each term, and do the same for the vertical gradient image

Now, add both images, and square root them This is our output image

22

||

∂∂

+

∂∂

=∇yG

xGG

Edge Detection – (5) How do we do this in MATLAB?

1) Create our Prewitt or Sobel Horizontal Masks: mask = fspecial(‘prewitt’); or mask = fspecial(‘sobel’);

2) Get the Vertical Mask: mask2 = mask’; ‘ transposes a matrix

3) Convolve with each mask separately dX = imfilter(im, mask); dY = imfilter(im, mask2);

4)Find the overall magnitude mag = sqrt(double(dX).^(2) + double(dY).^(2));

Edge Detection – (6) Note: We must cast images to double for sqrt() func. Now, we’ve found the overall gradient… how do we find

an actual edge? Threshold the image What do we mean by threshold?

Values greater than a threshold is an edge (white) Values less than the threshold are not edges (black)

How do we detect edges in MATLAB? Very easy! I = edge(im, ‘sobel’); or I = edge(im, ‘prewitt’);

So, we can pretty much ignore the previous slide, but I put that in there so you can see where it comes from

Edge Detection – (7) You can also call the routines this way: I = edge(im, ‘sobel’, thresh); or I = edge(im, ‘prewitt’, thresh); I specifies edges found in input (B & W image) im is the input image with edges to be found Second parameter specifies method of finding edges thresh determines threshold for finding edges If not specified, threshold will be found automatically

How do we use the thresh variable? Choose an intensity (e.g. 128) Any gradient value > 128 will be labeled white

Edge Detection – (8) Any gradient value < 128 is labelled black thresh is between [0,1], so take your threshold and

divide by 255 to use i.e. I = edge(im, ‘prewitt’, 128/255);

Output image will give you a black and white image White is an edge, black is not an edge Only works for B & W images. To find edges for colour

images, convert the colour image to a B & W image Use gray = rgb2gray(im); rgb2gray converts a colour image into B & W by

doing: I = (R + G + B) / 3 Each colour pixel is the average of the red, green and blue components

Edge Detection – (9) Sidenote: Here’s another way of finding the gradient Unsharp Masking

Probably heard this terminology in CSI? So what’s unsharp masking?

Let’s go back to Signals and Systems Suppose we perform a low-pass filtering of an image

Get a blurred version If we take the original image, and subtract its blurred

version, what are we doing? Removing all low frequency components, so what’s left? High frequency!

Edge Detection – (10) High frequencies are essentially edge information

Edges are sudden jumps Essentially high frequency How do we perform unsharp masking?

1) Blur the image Try using the following mask: mask = fspecial(‘average’, 5); %5x5

So… we blur! im_blur = imfilter(im,mask); 2) Subtract the original image from the blurred im_unsharp = im – im_blur;

Now, why are edges useful? We can use them to sharpen images Will take a look after this demo

Demo Time #6! Edge Detection in Images

Calculating an Unsharp Mask

Image Sharpening Last image enhancing topic Image Sharpening What is image sharpening? Make the image look

“clearer”, “sharper”, make the details “pop out more” How do we do this?

Edges are the “detail” behind the image Edges are high frequency

How do we make images sharper? Find overall magnitude of the gradient for the image Add these values on top of the original image Result? Increase the visibility of the edges Sharper

Image Sharpening – (2) Use imfilter with unsharp masking mask = fspecial(‘unsharp’);

The above syntax creates an unsharp mask in such a way where when you convolve, it will automatically subtract the image with a blurred version of itself and add the original image on top

Good for both x and y direction changes How? mask = fspecial(‘unsharp’); sharp = imfilter(im, mask);

This will perform the detection of abrupt changes, and adding them on top of the image all in one go.

Demo Time #7! Image Sharpening in Images

Segmentation via Thresholding There are times when we want to separate objects (the

things we want) from the background Segmentation To make this easier, we convert to a B & W image first if the

image is in colour. Else just leave it alone After conversion, most of the time the intensities of the

objects are distinctly different from background We can write code to save the pixels that match these

intensities, and disregard the rest Pixels that match Set to white, else set to black Output image will be binary White belonging to object,

and black belonging to background We can use this to mask out the background pixels

Segmentation via Thresholding (2) What do we do?

Take a look at the histogram, and see which pixels are predominantly belonging to the object, and of the background

We threshold the image using this information We then use this threshold map to figure out what pixels

we want to keep, and what we want to disregard

Demo Time #8! Object Segmentation via Thresholding

Noise Filtering A common task in image processing is to eliminate or

reduce image noise from an image Image Noise: Pixels in an image that are corrupted

undesirably Pixels could be corrupted in the acquisition process, or

in transmitting, etc. How do we reduce image noise?

Think in the frequency domain Noise is essentially high-frequency information

If we blur the image, we would eliminate the noise, but the quality would reduce blurring details

Noise Filtering – (2) We can generate artificial noise and add these to

images Purpose is for research Design good filters by

recreating the noise we would encounter in practice How do we generate artificial noise? Use imnoise

We will be concerned with two ways of generating noise out = imnoise(im, ‘gaussian’, mean, var);

out = imnoise(im, ‘salt & pepper’, prob);

First method generates Gaussian-based noise Usually encountered in transmitting process

Noise Filtering – (3) Second method generates “salt & pepper” based noise

Also known as impulsive noise Called this way, because for monochromatic images, it

literally looks like someone took salt (white pixels) and pepper (black pixels) and shook it over the image

For colour images, specks of pure red, green and blue pixels appear

We can use blurring to get rid of Gaussian noise, but for impulsive noise, we need to use median filtering What’s median filtering? Like convolution, but we’re not

doing a weighted sum

Noise Filtering – (4) For each pixel (r,c) in the image, extract an M x N

subset of pixels centered at (r,c) Sort these pixels in ascending order, and grab the

median value The output image at (r,c) is this value

How do we perform median filtering in MATLAB? out = medfilt2(im, [M N]);

Demo Time #9! Noise Filtering in Images

Template Matching Template matching is using a small test image, which

we’ll call a patch Objective is to automatically find the location of where

this patch is in the entire image Useful in a variety of applications: Image Retrieval, Eye

Detection, etc. What I’ll show you today is to do some basic image

stitching We will have two images of the same scene, that have

been taken at slightly different perspectives Only horizontal shifting is concentrated on here

Template Matching – (2) What’s the best way to do template matching?

1) For each pixel (r,c) in the image, extract a subset that is the same size as the template with its centre at (r,c)

2) Perform a cross-correlation between the patch and this subset

3) Take this value and assign it to location (r,c) for the output

4) The best location of where the template is, is where the maximum cross-correlation is

We can perform (1) to (3) by doing: C = normxcorr2(template, im);

Template Matching – (3) Output is a correlation map

To find the row and column co-ordinates of where the template best matches, do the following:

[row col] = find(C == max(C(:)));

So, how do we do image stitching? 1) Extract a region in either image that is common

between both Template 2) Find the co-ordinates of where this template is in

both images 3) Determine how much vertical and horizontal

displacement there is between the two images

Template Matching – (4) 4) Create an output image with dimensions that

encapsulate both images together 5) Place one image on the left, then place the other

image by displacing it over by the horizontal and vertical shift

OK… let’s do some code!

Demo Time #10! Simple Image Stitching

Conclusion MATLAB is a great tool for digital image processing Very easy to use This is not an exhaustive tutorial! There are many

more things you can do with MATLAB For more image processing demos, check out:

http://www.mathworks.com/products/image/demos.html Lots of cool image processing stuff you can find here

For a more comprehensive MATLAB tutorial, check: http://www.ee.ryerson.ca/~rphan/ele532/MATLABTutorial.ppt

You can access the slides, images and code at: http://www.rnet.ryerson.ca/~rphan/IEEEDIPTalk

http://www.mathworks.com/products/image/demos.html


http://www.rnet.ryerson.ca/~rphan/IEEEDIPTalk

introduction to digital image processing using matlab

Engineering

digital images color

actual images

matlab quick intro

digital images simple

image histograms intro

lot easier matlab

andor use matlab

comprehensive matlab