albertofernandez's profile - activity

I am reading a paper. It says like this:

"For experiments conducted on noisy images, each texture image was corrupted by additive Gaussian noise with zero mean and standard deviation that was determined according to the corresponding Signal-to-Noise Ratios (SNR) value."

And then, they show the classification rate (%) on UIUC database with additive gaussian nosie of different Signal-To-Noise Rations (SNR):(SNR=100 SNR=30 SNR=15 SNR=10 SNR=5)

So I want to do the same....

Is GaussianBlur my function? How do I determine the SNR?

In connection with the image: an 8-bit single-channel image. Non-zero pixels are treated as 1's. Zero pixels remain 0's,

I do not understand you very well, but I think that you should try this: res = cv2.bitwise_and(img, img, mask=imgf)

I do not understand you very well, but I think that you should try this: cv2.bitwise_and(img, img, mask=imgf)

x_centroid = round(M['m10'] / M['m00']) y_centroid = round(M['m01'] / M['m00'])

x_centroid = round(M['m10'] / M['m00']) y_centroid = round(M['m01'] / M['m00'])

x_centroid = round(M['m10'] / M['m00']) y_centroid = round(M['m01'] / M['m00'])

x_centroid = round(M['m10'] / M['m00']) y_centroid = round(M['m01'] / M['m00'])

This python code performs what you want. # Import required packages: import cv2 # Load the image and convert it to gr

Many thanks @sturkmen :). It's nice to ear things like this

Images are numpy arrays. Therefore you can use numpy capabilities to create an image contaning two images. The images to

Deteccion de persona utilizando python- openCV. I want to detect the movement of a person using: bodydetection = cv2.Ca

Deteccion de persona utilizando python- openCV. I want to detect the movement of a person using: bodydetection = cv2.Ca

There are a lot of papers comparing Local Binary Pattern (LBP) versus Local Ternary Pattern (LTP), or modifications to the original LBP operator like Center-symmetric local binary patterns (CS-LBP), Local quinary patterns (LQP), Completed Local Binary Pattern (CLBP) and so on.

All this methods belong to the same type and a framework for texture analysis can be build. This framework for texture analysis is called HEP (histogram of equivalent patterns) and it is described in this paper: Texture description through histograms of equivalent patterns

This is the project web page, with Matlab implementation of hep (hep.m):

http://dismac.dii.unipg.it/hep/index.html

It would be fantastic if Opencv had HEP implementation.

I have written some lines of pseudo-code assuming there is an Opencv implementation:

//create a hep descriptor. In this case, a uniform local binary pattern descriptor for gray-scale images 
int neighbors = 8;
int radius = 1;
//char is for    values between 0...255 (gray images)
HEP<LBP<u2>,char> * hep_lbp_u2_descriptor = new HEP<LBP<u2>>(neighbors,radius,..); 

//load a gray-scale image to test the descriptors
Mat image = imread(...., );

Mat gray_image;
cvtColor( image, gray_image, CV_BGR2GRAY );

 //1. First, you could have the possibility to test a "raw pattern" in order to 
 //see how this descriptor works for a given pixel

 Mat image_test = Mat::zeros(Size(3, 3), CV_8UC1);
 image_test.at<uchar>(0,0) = 1;
 image_test.at<uchar>(0,1) = 2;
 image_test.at<uchar>(0,2) = 3;
 image_test.at<uchar>(1,0) = 4;
 image_test.at<uchar>(1,1) = 4;
 image_test.at<uchar>(1,2) = 6;
 image_test.at<uchar>(2,0) = 7;
 image_test.at<uchar>(2,1) = 8;
 image_test.at<uchar>(2,2) = 9;

 //compute the "raw pattern" for the central pixel (1,1)
 cout<<"the value for the central pixel is"<<hep_lbp_u2_descriptor->compute_raw_pattern(image_test,1,1)<<endl;

 //2. You could use this descriptor to build a lbp-image (a hep-image):
 Mat lbp_image;
 hep_lbp_u2_descriptor->compute_map(gray_image,lbp_image);

 //3. You could have the posibility to build the lbp-histogram (a hep-histogram):
 Mat lbp_histogram;
 int width_divisions = 5;
 int height_divisions = 6;
 hep_lbp_u2_descriptor->compute_histogram_grid(gray_image, lbp_histogram, width_divisions ,height_divisions );
 //(in this case the lbp_histogram would have 59*5*6 features 
 //(uniform lbp with a neighborhood of 8 pixels))

Based on comment from this question, and coments from the other related question (Fire/Flame Detection using OpenCV):

@StevenPuttemans: "Another way could be to use the fact that fire actually involves motion between frames in a video input"

@Guanta "You could train a cascade-classifier with LBPs since LBPs are also often used for texture detection/recognition this would be worth trying"

@red-viper
"is there any other way that uses machine learning?"

@pklab "The example shows that in the visible spectrum, the flame visibility depend on context. In the IR, the context is less relevant. After this you could use ML on IR image too."

I would like to complete the proposed hardware answer with a software one that i think could work together.

In this paper: "Early Fire Detection Using HEP and Space-time Analysis" [1] motion (as suggested by @stevenputtemans) and texture analysis (as suggested by @guanta) are involved.

You could also check "Video Fire Detection - Review" [2], Real-time Fire Detection for Video Surveillance Applications using a Combination of Experts based on Color, Shape and Motion [3], Automatic fire pixel detection using image processing: A comparative analysis of Rule-based and Machine Learning-based methods [4]

[1] http://arxiv.org/pdf/1310.1855.pdf

[2] http://signal.ee.bilkent.edu.tr/Publi...

[3] https://www.researchgate.net/profile/...

[4] https://hal.archives-ouvertes.fr/hal-...

I agree with @StevenPuttemans

In order to get more images, you can use GENKI-4k Face, Expression, and Pose Dataset [1]

The GENKI-4K dataset contains 4,000 face images spanning a wide range of subjects, facial appearance, illumination, geographical locations, imaging conditions, and camera models. All images are labeled for both Smile content (1=smile, 0=non-smile) and Head Pose (yaw, pitch, and roll parameters, in radians).

I have used this database to train a smile classifier using LBP + SVM approach (some years ago) and results were quite satisfactory. Steps:

• (1) Face detection
• (2) Get Local Binary Pattern from the face
• (3) SVM classification

Step (2) can be decomposed in:

• crop face region
• normalize face region based on eye coordinates
• histogram equalization or something similar
• Apply LBP to a NxM non-overlapping region

The public GENKI-4K dataset is available for download here [2].

[1] http://mplab.ucsd.edu/wordpress/?page...

[2] http://mplab.ucsd.edu/wordpress/wp-co...

Two sample images:

• Face detection
• HSV conversion
• Small rectangle region below the detected face is calculated
• Calculate dominant color type in the region --> getPixelColorType function (color types: ("Black", "White","Grey","Red","Orange","Yellow","Green","Aqua","Blue","Purple","Pink"))
• In order to calculate different types of clothes, you can compute LBP features over this small rectangle (or more regions) because LBP features can discriminate textures quite well.

References:

shirtDetection