uib-vfeatures 0.5.1

Vision features of generalistic use

UIB - V Features

Introduction

UIB - V Features provide a set of useful features. With three types of features: morphological, texture and color. All the features can be used with mask or with the contours. Every feature is a numerical value that used in ML can improve their results.

The morphological features are all grouped in one iterator, so you can calculate all the features inside a loop easily.

The library use mask and contours. Masks are two value image, where the object has a value diferent than the rest of the image. A contour is a 2D vector of points that define a contour. To calculate a contour normally is used the OpenCV function.

Installation

Install the library is very simple with pip

pip install uib-vfeatures

---

List of features

Morphological

#####All this features are in the iterator

• Solidity
• Convex hull perimeter
• Convex hull area
• Bounding box area
• Rectangularity
• Minor radius
• Maximum radius
• Feret
• Breadh
• Circularity
• Roundness
• Feret Angle
• Eccenctricity
• Center
• Sphericity
• Aspect Ratio
• Area equivalent diameter
• Perimeter equivalent diameter
• Equivalent elipse area
• Compactness
• Area
• Convexity
• Shape
• Perimeter

Color

• Mean of the LAB channels
• Mean of the RGB channels
• Mean of the HSV channels
• Standard deviation of the LAB channels
• Standard deviation of the RGB channels
• Standard deviation of the HSV channels

Textures

The texture features depends on the parameter of a unique function. The first two parameter define the texture, with the distance and the angle of the texture. The third defines the feature to extract and the last one is a grey-scale image.

Texture features

• Contrast
• Dissimilarity
• Homogeneity
• ASN
• Energy
• Correlation

---

Demo

We're going to use our library with a mask image .

from uib_vfeatures.masks import Masks
from uib_vfeatures import Features_mask as ftrs
import cv2

First of all we read the image from a file, then we try our features with visualizations. We only have three features with visualization: the bounding box area, the eccentricity and the solidity.

mask = cv2.imread("mask.jpg")

Masks.bounding_box_area(mask, True)

Masks.eccentricity(mask, True)
Masks.solidity(mask, True)

Iterator

You can use an iterator and implement every morpholical feature.

features = {}

for key, func in features.items():
    features[key] = func(mask)

As a result we had a dicctionary of the form {'Feature_name': value}