Color-Quantization-and-Circle-Detection-

Computer Vision Assignment on Color Quantization using K-means and Circle Detection using Hough Transform

The focus of the Repository:

This Repository contains a code solutions along with a Report for a Computer Vision Assignment for two problems:

• Quantization of Color Space using K-means Clustering:
• Implementation of Hough Transform Circle Detector:

Problem - 1: COLOR QUANTIZATION:

---

Given Image:



• Quantize a 3-dimensional RGB image and map pixels to its nearest k-means center
• Converting a 3-dimensional RGB image into a HSVand quantize the 1-dimensional hue space
• Compute the SSD error between RGB image pixels and quantized pixels for both RGB and HSV space
• Displaying the histograms of hue values with equally spaced bins and bins defined by K cluster centers

Results

RGB Quantized



HSV Quantized



Histograms





SSD Error





Problem - 2: CIRCLE DETECTION

---

Given Image:




• Implementing a hough transform Circle Detector that takes an input image and fixed radius and returns centers of any detected circles.
• The output centers is a NX2 matrix in which each row lists (x,y) position of the detected circles' center
• Displaying the Hough Transform accumulator array
• Experimenting on how to determine the number of circlepresent by post-processing the accumulator array
• Describing the impact of the vote space quantization(bin size)/i>

Results

Detection Radius =10 and accumulator array




Detection Radius =30 and accumulator array




Detection Radius =100 and accumulator array




Repository Files:


Code Implementation:


• ColorQuantize.py - main code that calls the other helper functions
• computeQuantizationerror.py - SSD error between quantized and original pixels
• quantizeHSV.py - converts RGB image to HSV and quantize 1-D Hue space
• quantizeRGB.py - quantizes 3D RGB image with K-means
• detectCircles.py - code implementation for CircleDetection

Image png files:


• Image 1 - fish.png
• Image 2 - jupiter.png

Report