Overview

Dominant colors in an image are detected according to their Hue, Saturation and Value of intensity. This is extended to videos by applying the same to each frame. For video purpose, I’ve used webcam. This project is done using OpenCV, a package for Image Analysis in Python.

Requirements

Install Python3.x, and then install OpenCV by following this page. That page is escpecially the installation instructions for mac users.

Concept

• First, HSV format is extracted from tthe original image/frame.

  hsv_version = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
  

• We make the filtering using single mask layer that we create. Resultant will be the bitwise_and of the mask and the original frame. Let’s see the mask later, I’ve the stub below to explain.

  res = cv2.bitwise_and(frame, frame, mask=mask)
  

• Actually, the mask is not yet defined. Let’s see what the mask actually is. In OpenCV, Hue ranges from 0 to 180, Saturation tanges from 0 to 255 aand Value of Intensity ranges from 0 to 255. Below showed picture describes the variation.

• Once we get these upper and lower bounds, we can create a mask allowing pixels which fall in these bounds.

  mask = cv2.inRange(hsv_version, lower_bound, upper_bound)
  

• For, smoothening the errors, the false positives and true negatives. False positives are the values included which should not be. Whereas, true postives are those excluded from considering. This process is sometimes referred to as averaging.

  kernel = np.ones((10,10), np.float)/100
  smoothened = cv2.filter2D(res, -1, kernel)
  

Results

Results without the smoothening filter, resultant with smoothening filter, and the original, all are presented as output.