Color, Vision, Image Processing Notes

Color Appearance Term : Colorfulness vs. Chroma

2015-07-26T09:23:00.000-07:00

With intuitive example, in this video I have explained the what is colorfulness, what is chroma, and what are the differences between colorfulness and chroma.

Can Apple buy Greece?

2015-07-21T12:38:00.001-07:00

If you had enough money, would you buy Greece? Apple does, should Apple buy Greece? What do you think?

From Color Appearance point of view, what is the difference between "Brightness" and "Lightness"

2015-07-20T15:30:00.002-07:00

In color science, brightness and lightness are related, but two different things. Let us discuss that in the following video!

Color Constancy Intuitively Explained !

2014-08-27T06:25:00.002-07:00

Color Constancy Intuitively Explained !

Benefit of Intrinsic Images

2014-08-07T02:47:00.000-07:00

Intrinsic images refers to decomposition of an image into illumination S(x), reflectance (or albedo) R(x) and specular components C(x) [1].

I (x) = S(x) R(x) + C(x)

But, what are benefits of such decomposition?
-----------------------------------------------
1. Shape from Shading -- Just focus on the shading effect,that is, illumination component
2. Segmentation -- separating cast shadows from image

Notes
----------
1. Estimating the decomposition is a ill-posed problem since for every observed pixel, we have multiple unknowns.

Reference :
[1] Grosse, Roger, et al. "Ground truth dataset and baseline evaluations for intrinsic image algorithms." ICCV 2009.

Simple Boxplot where each five boxplots are clustered together

2014-07-29T13:04:00.000-07:00

a=rand([20 25]);
positions = [1:0.5:3 4:0.5:6 7:0.5:9 10:0.5:12 13:0.5:15];
boxplot(a,'positions', positions)

More side by side Boxplot

2014-07-29T02:28:00.000-07:00

Apart from my previous post , another approach for side by side Boxplot is the following!

Side by Side boxplot in MATLAB

2014-07-29T01:35:00.001-07:00

I have found the following code somewhere in the internet but forgot to keep track of the link. Anyway, its for you!

See below the MATLAB code that generates this figure!

Convert RGB to CIE LAB with explanation, MATLAB example

2014-07-26T04:50:00.000-07:00

In this tutorial. I have explained what is RGB and CIE LAB color space. Then, I have shown an example in MATLAB where I converted an image from sRGB to LAB

http://youtu.be/kB2q0thrCTw

Excel/Powerpoint Figure to EPS(ultimately Figure in PDF) for LATEX

2014-07-22T16:25:00.000-07:00

This video shows how to convert a figure in Excel/OpenOffice into a eps format (vector graphics) so that you may use it in latex

Calculating : Pearson Correlation Coefficient (without Statistics toolbox)

2014-07-18T06:27:00.002-07:00

This is the equation to compute Pearson Correlation Coefficient between x and y variable. In other words, if x and y are column vector and we want to know how much they are correlated, we can use the above formula find our r. If r is close to 1, its positively correlated if negative 1 its negatively correlated so on and so forth. See the MATLAB code snippet below. You can compare the result with matlabs implementation if you have Statistics Toolbox

How to Create Zero Mean white Gaussian Noise in MATLAB

2014-07-15T02:35:00.001-07:00

You can create zero mean Gaussian Noise using the randn function. It is also known as white noise.
See the following script.

Compute Black body illuminants

2013-07-24T02:04:00.001-07:00

Using the black body radiation concept, we can compute XYZ of all black-body illuminant. Prediction of this illuminants seems to be very close to CIE recommended illuminants such as CIE illuminant A

Let's show a quick example. We know illuminant A 2856K has 10 degree observer based X =111.144, Y=100, Z=35.200

Now, lets compute using our black body illuminant formula

get_XYZ_of_blackbody(2856).*100

ans =

111.1385 100.0000 35.2127

Pretty close! :)

Have fun!

How to convert RGB to XYZ or vice versa

2013-07-14T09:54:00.001-07:00

How to convert RGB to XYZ or vice versa?

Conversion matrices can be found here
http://www.brucelindbloom.com/index.html?Eqn_RGB_XYZ_Matrix.html

% You need MATLAB or Octave to run this

% Let say you have one color and its pure red

red=[255;0;0]

% now, you can convert it to XYZ assuming rec 709 display and D65 white

% point using the following matrix

RGB2XYZ=[0.4124564 0.3575761 0.1804375 0.2126729 0.7151522 0.0721750 0.0193339 0.1191920 0.9503041]; redXYZ=RGB2XYZ*red;

%now, you can normalize it further so that Y=100 r

edXYZ=redXYZ/redXYZ(2)*100;

% this is the final result

redXYZ = 193.9393 100.0000 9.0909

Compute Trimean from Quantile in Matlab

2013-07-11T02:30:00.001-07:00

Since, Trimean is robust to outliers, it is recommended as summary statistics instead of simple mean. Let us give an example.

x = [2 5 6 10 11 13 200];

In this example, taking a mean is heavily influenced by outlier value 200

mean(x)

ans =

35.2857

But is there any better summary statistics than taking just direct mean?

Q = quantile(x,[0.25, 0.5, 0.75])

Q =

5.2500 10.0000 12.5000

Note that here Q(2) is the median. And, Q(1) and Q(3) are first and third quantile respectively.

Now, we can compute the trimean by the weighted sum of quantiles. Median gets the higher weight since its the most robust quantile among three.

0.25*Q(1)+0.5*Q(2)+0.25*Q(3)

ans =

9.4375

This trimean represent x which is a better summary statistics than simple mean.

Normalize Ellipse into Circles by Affine transformation

2013-07-02T15:45:00.000-07:00

The following code requires MexOpenCV to be installed in your system

Original Image

One example of warpedIgray

After one pass (when i=1)

Captured Ellipse from warpedIgray

Normalized to circles by affine transformation

Extract MSER and Fit ellipses using OpenCV (MexOpenCV)

2013-07-01T13:40:00.004-07:00

For this small tutorial make sure you have MexOpenCV installed in your system.

We will first extract maximally stable extremal regions (MSER). Second, for each MSER we will fit an ellipse. Finally, we will draw those ellipses on the image.

All MSERs are fitted to ellipses!

Fitted Ellipses are drawn on the source image

To view a particular ellipse captured from the image after positioning in a up right situation

Rotated according to angle of rectangle covering the ellipse. Please see the above code

After rotation, image patch is extracted using getRectSubPix

now, resize the above patch into a fixed size patch such as 30x30 square

resized to 30x30 square using cubic interpolation

cv.fitEllipse calculates the ellipse that fits (in a least-squares sense) a set of 2D points best of all. It returns the rotated rectangle in which the ellipse is inscribed. The algorithm [Fitzgibbon95] is used.

[Fitzgibbon95] https://homepages.inf.ed.ac.uk/rbf/MY_DAI_OLD_FTP/rp810.pdf

http://answers.opencv.org/question/497/extract-a-rotatedrect-area/

Image Bulk resizing using Picasa

2013-07-01T06:11:00.002-07:00

Click the Folder to select (Containing images to be resized)

FIle->Export Picture to Folder

In word 2007 how to fixate a text box?

2013-07-01T03:16:00.001-07:00

Select the text box (or drawing canvas)
Format->position->More Layout Position

Uncheck Move object with text
Check Lock anchor
Uncheck Allow overlap
Check layout in table cell

Excel figure into Word

2013-06-19T09:38:00.000-07:00

Copy the figure and paste it as Microsoft Office Graphic Object (it works like EPS!)

Convert PDF to EPS

2013-06-13T15:07:00.002-07:00

1. Open the PDF by GSview (Download if not installed already in your system)
2. In GSView: File->Print->Post Script Printer->OK, Make sure that Print to File is checked

PowerPoint to PDF or EPS to CROP to ......

2013-06-13T13:50:00.001-07:00

I was on my way to become crazy in order to convert my power point figure into a cropped version of eps or pdf!

Probably Simpler option is to install foxit and do the following
1. Home->Snapshot->
2. File->Create from Clip Board-> Save PDF
Use in latex \includegraphics[height=1.5in]{images/top_cropped}

Another Approach:

Here is what you should do step by step

1. Make sure Ghost Script, GIMP is installed in your system!
2. First convert your powerpoint (or whatever it is) into PDF. Now, the problem is its too big and there might be unnecessary part in the pdf/eps which you would like to discard.
3. Open the PDF using GIMP. Then Image-->Zealous Crop (or Crop Manually using select tool)
4. File->Export (Choose your format like EPS/PDF etc)

Good luck future me!

watch out while using sRGB2lab

2013-03-12T14:36:00.000-07:00

Imagine you have a 8 bit/color channel (24bit image) and you want to convert RGB values (assuming in sRGB color space) into LAB color space. SO you do the following

rgb=imread('peppers.png');
lab = applycform(rgb,makecform('srgb2lab'));

Where rgb is a NxMx3 matrix of unint8 type and so does the lab. Everything seems fine! But try the following:

Lmax= max(max(lab(:,:,1))) % maximum L of LAB should be within 100

but is it the case? NO?

You will notice that Lmax is beyond 100 and the reason is lab is in uint8 type

Therefore, if you want a normalized L value from LAB do the following

lab_d=lab2double(lab); % normalization !!!

Now try

max(max(lab_d(:,:,1))) % L should be <=100

Therefore, watch out! But, color your life! :)

simplest way to use EXR in MATLAB

2013-01-29T03:00:00.001-08:00

How to read/write EXR images in MATLAB?

1. You can Install HDRITOOLS from here (https://bitbucket.org/edgarv/hdritools/downloads); Choose *x64.msi or *.x86.msi based on your current MATLAB installation

2. Say install location of HDRITOOLS is ~Documents\MATLAB\HDRIToolsx64\matlab, here please try reading a EXR file using following command

info=exrinfo('A003C001_120626W700021.exr');

info.channels

info.channels(1) % prints out the name of the first channel

info.channels(2) % prints out the name of the second channel

info.channels(3) % prints out the name of the third channel

c1=exrreadchannels( 'A003C001_120626W700021.exr',info.channels(1) ); % reading first channel

Finding a vector within a matrix

2013-01-23T08:45:00.000-08:00

Imagine, I have a matrix called RGB (list of some colors) like the following

RGB=[1 2 3; 4 5 6; 7 8 9; 10 11 12; 0 0 0; 1 2 5]

RGB =

1 2 3
4 5 6
7 8 9
10 11 12
0 0 0
1 2 5

I would like to find whether vector [ 1 2 5] is available in RGB or not?

The fastest way is to do some logical operation.

loc = (RGB(:,1)==1 & RGB(:,2)==2 & RGB(:,3)==5 )

loc =

0
0
0
0
0
1

So now we know where is the vector [1 2 5] is located. We can find its index or even how many times it occurs by manipulating above loc vector