Image Retrieval: Color Coherence Vector

Updated on September 30, 2018
TarekMamdouh profile image

Tarek is a software engineer, enjoys implementing side research projects and blogging them.

Content-Based Image Retrieval System
Content-Based Image Retrieval System

Introduction and a brief literature history

Content-based image retrieval, is the field that is concerned with being able to retrieve an image based on the actual content of it (not based on any textual/meta data attached with it). The process of retrieving the right features from the image is done by an image descriptor. One important use case for any image descriptor is the ability to use its generated features to define the similarity between images

In this post, we are going to talk about one of the commonly known techniques used in image retrieval which is Color coherence vector, it is an image descriptor (or more specifically, it is a color descriptor), that extracts color-related features from the image which can be used as a low dimensional representation of this image.

Global Color Histogram (GCH) and Local Color Histogram (LCH). Both descriptors are based on computing the Color Histogram of the image, the difference is that GCH computes the color histogram for the whole image and uses this frequency table as a low dimensional representation of the image, while on the other hand, LCH first partitions the image into blocks and each block will have a separate color histogram calculated, and the concatenation of these local color histograms is the low dimensional representation of the image.

Due to the sparsity of the resultant color histogram representation, some papers (like "Local vs. Global Histogram-Based Color Image Clustering") suggest applying Principle Component Analysis (a method used for dimensionality reduction, and extracting only the useful features) to the outputted color histograms.

However, these methods have some clear issues, for example GCH doesn't encode any information about the color spatial distribution in the image. LCH performs much better than GCH as it overcomes this specific problem to some extent, but it is still not robust enough to some little variations like image rotations and flips.

Now, we’ll discuss a more useful yet fast color descriptor that is capable of encoding information about color spatial distribution which is called Color Coherence Vector (CCV).

Color Coherence Vector

Color Coherence Vector (CCV) is a more complex method than Color Histogram. It works by classifying each pixel as either coherent or incoherent. Coherent pixel means that it is part of a big connected component (CC) while incoherent pixel means that it is part of a small connected component. A crucial step for this method to work is defining the criteria by which we decide whether a connected component is big or not.

How features are extracted in CCV?

These steps target building a low dimensional representation of the image.

  1. Blur the image (by replacing each pixel’s value with the average value of the 8 adjacent pixels surrounding that pixel).
  2. Quantize the color-space (images’ colors) into n distinct color.
  3. Classify each pixel either as coherent or incoherent, this is computed by
    • Finding the connected components for each quantized color.
    • Determining the tau’s value (Tau is a user-specified value, normally, it’s about 1% of image’s size), any connected component with number of pixels more than or equal to tau then its pixels are considered coherent otherwise they are incoherent.
  4. For each color compute two values (C and N).
    • C is the number of coherent pixels.
    • N is the number of incoherent pixels.

    It’s clear that the summation of all colors in C and N should be equal to the number of pixels.

Let’s take this example to concretely describe the steps of algorithm.
Assuming that the image has 30 unique colors.

Now we’ll quantize the colors to only three colors (0:9, 10:19, 20, 29). This quantization is essentially about combining similar colors to a single representative color.

Assuming that our tau is 4

For color 0 we have 2 CC (8 coherent pixels)

For color 1 we have 1 CC (8 coherent pixels)

For color 2 we have 2 CC (6 coherent pixels and 3 incoherent pixels)

So finally our feature vector is

Defining a distance function

The purpose of having a distance function is to quantify the dissimilarity between any two images. It complements the usefulness of the color descriptor, for example, the color descriptor can extract features for all images and store them in a database and then during the image retrieval phase this distance function will be used to retrieve the image with minimum distance to the original query image.

In order to build a distance function for CCV, we use the calculated coherent and incoherence features (C and N for each color) in our distance function to compare between any two images (let's name them a and b, in the following equation).

Ci : number of coherent pixels colored with i.

Ni : number of incoherent pixels colored with i.

Drawbacks of Color Coherence Vector

Now we see that Color Coherence Vector method considers information about color spatial distribution between pixels in its coherence component. But this method has some drawbacks. The remaining part of this post will discuss two main drawbacks of it.

Coherent pixels in CCV represent the pixels which are inside big noticeable components in the image. However, if we combined these entire components into one component, we will end up having only one bigger component where the number of its pixels will be equal to the number of the pixels in the two original big components.

To make it clear, let's look at these pictures (assuming tau equals to 8).

Although they are different pictures but they have the same CCV.

It might be clear that this problem could be solved by adjusting the threshold tau, but still tuning it is not trivial, because in many cases you will need to choose between multiple thresholds, each one of them is still not completely correctly capturing the difference between big components and small ones in you image dataset.

Another problem we may encounter is the positions of these remarkable connected components relative to each other.

The following pictures have the same CCV but with different appearance:

There are many solutions to this problem. For example, adding another dimension in the feature vector that would capture the components’ position relative to each other may break these ties. This paper "An Improved Color Coherence Vector Method for CBIR" describes this approach.

Here is the link of CCV paper in case you would like more academic details description of the method. I hope this post was beneficial to you, lastly, you can find my Matlab implementation of CCV on Github (ColorCoherenceVector Code).

Questions & Answers

    © 2013 Tarek Mamdouh


      0 of 8192 characters used
      Post Comment
      • profile image

        muhammad junaid 

        10 months ago

        sir please tell me that how to test the similirity percentage of two image according to color coherence vector,,

        thank you..


      This website uses cookies

      As a user in the EEA, your approval is needed on a few things. To provide a better website experience, uses cookies (and other similar technologies) and may collect, process, and share personal data. Please choose which areas of our service you consent to our doing so.

      For more information on managing or withdrawing consents and how we handle data, visit our Privacy Policy at:

      Show Details
      HubPages Device IDThis is used to identify particular browsers or devices when the access the service, and is used for security reasons.
      LoginThis is necessary to sign in to the HubPages Service.
      Google RecaptchaThis is used to prevent bots and spam. (Privacy Policy)
      AkismetThis is used to detect comment spam. (Privacy Policy)
      HubPages Google AnalyticsThis is used to provide data on traffic to our website, all personally identifyable data is anonymized. (Privacy Policy)
      HubPages Traffic PixelThis is used to collect data on traffic to articles and other pages on our site. Unless you are signed in to a HubPages account, all personally identifiable information is anonymized.
      Amazon Web ServicesThis is a cloud services platform that we used to host our service. (Privacy Policy)
      CloudflareThis is a cloud CDN service that we use to efficiently deliver files required for our service to operate such as javascript, cascading style sheets, images, and videos. (Privacy Policy)
      Google Hosted LibrariesJavascript software libraries such as jQuery are loaded at endpoints on the or domains, for performance and efficiency reasons. (Privacy Policy)
      Google Custom SearchThis is feature allows you to search the site. (Privacy Policy)
      Google MapsSome articles have Google Maps embedded in them. (Privacy Policy)
      Google ChartsThis is used to display charts and graphs on articles and the author center. (Privacy Policy)
      Google AdSense Host APIThis service allows you to sign up for or associate a Google AdSense account with HubPages, so that you can earn money from ads on your articles. No data is shared unless you engage with this feature. (Privacy Policy)
      Google YouTubeSome articles have YouTube videos embedded in them. (Privacy Policy)
      VimeoSome articles have Vimeo videos embedded in them. (Privacy Policy)
      PaypalThis is used for a registered author who enrolls in the HubPages Earnings program and requests to be paid via PayPal. No data is shared with Paypal unless you engage with this feature. (Privacy Policy)
      Facebook LoginYou can use this to streamline signing up for, or signing in to your Hubpages account. No data is shared with Facebook unless you engage with this feature. (Privacy Policy)
      MavenThis supports the Maven widget and search functionality. (Privacy Policy)
      Google AdSenseThis is an ad network. (Privacy Policy)
      Google DoubleClickGoogle provides ad serving technology and runs an ad network. (Privacy Policy)
      Index ExchangeThis is an ad network. (Privacy Policy)
      SovrnThis is an ad network. (Privacy Policy)
      Facebook AdsThis is an ad network. (Privacy Policy)
      Amazon Unified Ad MarketplaceThis is an ad network. (Privacy Policy)
      AppNexusThis is an ad network. (Privacy Policy)
      OpenxThis is an ad network. (Privacy Policy)
      Rubicon ProjectThis is an ad network. (Privacy Policy)
      TripleLiftThis is an ad network. (Privacy Policy)
      Say MediaWe partner with Say Media to deliver ad campaigns on our sites. (Privacy Policy)
      Remarketing PixelsWe may use remarketing pixels from advertising networks such as Google AdWords, Bing Ads, and Facebook in order to advertise the HubPages Service to people that have visited our sites.
      Conversion Tracking PixelsWe may use conversion tracking pixels from advertising networks such as Google AdWords, Bing Ads, and Facebook in order to identify when an advertisement has successfully resulted in the desired action, such as signing up for the HubPages Service or publishing an article on the HubPages Service.
      Author Google AnalyticsThis is used to provide traffic data and reports to the authors of articles on the HubPages Service. (Privacy Policy)
      ComscoreComScore is a media measurement and analytics company providing marketing data and analytics to enterprises, media and advertising agencies, and publishers. Non-consent will result in ComScore only processing obfuscated personal data. (Privacy Policy)
      Amazon Tracking PixelSome articles display amazon products as part of the Amazon Affiliate program, this pixel provides traffic statistics for those products (Privacy Policy)
      ClickscoThis is a data management platform studying reader behavior (Privacy Policy)