The for loop simply iterates over all the colors retrieved from the image.įor each color, the loop changes it to lab, finds the delta (basically difference) between the selected color and the color in iteration and if the delta is less than the threshold, the image is selected as matching with the color. We use the method rgb2lab to convert the selected color to a format we can compare. We first extract the image colors using our previously defined method get_colors in RGB format. Finally, to combine paths while reading files from a directory, we import os. To compare colors we first convert them to lab using rgb2lab and then calculate similarity using deltaE_cie76. KMeans algorithm is part of the sklearn's cluster subpackage. To extract the count, we will use Counter from the collections library. We import the basic libraries including matplotlib.pyplot and numpy.

The sample_image.jpg was clicked by me and the other 5 images in the folder images were taken from Unsplash. The complete notebook is available at this repository.

In this article, I explain how I understood the basics of OpenCV, extracted colors from images using KMeans algorithm and filtered images from a collection of images based on RGB values of colors. I got inspired to actually write the code that can extract colors out of images and filter the images based on those colors. We’ve all seen that we can search online on the basis of certain filters one of which is color. When I came across OpenCV which allows import and manipulation of images in Python, I started to wonder if information could be extracted out of those images using Machine Learning and used in some way. I recently started reading about how I could work with Images in Python.

Color extract software