# Image Quality Assessment (IQA)

The IQA tool helps you analyze and clean up images. When selecting "IQA Generation" during Data Import, the system generates data for each image, including brightness, contrast, and more.

{% hint style="info" %}
**Hint**

You can choose whether to perform IQA when <mark style="color:blue;">**importing a dataset**</mark>.
{% endhint %}

## **Quick Start**

**Import data with IQA (image quality assessment).** Click on the video below for a quick overview:

{% embed url="<https://youtu.be/_-OnnNf14Jo>" %}

### IQA Items：

* Lightness： Overall lightness of an image
* Contrast： Overall contrast of an image
* Colorfulness： The perceived vividness of colors of an image
* Max Color Kurtosis： The fourth color moment
* BRISQUE： Blind/Referenceless Image Spatial Quality Evaluator
* Laplacian-Blur： Applying Laplacian operator and taking maximum over the filtered pixel values
* FFT-Blur： Use Fast Fourier Transform (FFT) to covert an image to the frequency domain

<figure><img src="https://2101974232-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FO2GXP74UOzykZuHBn8HP%2Fuploads%2FqA1F3nwNm5R7bZsTyIoQ%2Fimage.png?alt=media&#x26;token=80d4be88-1e39-404d-9cd0-d5022cd01efe" alt=""><figcaption></figcaption></figure>

Use Data Visualization to filter data and apply default image parameters to remove problematic images.

Clicking on Metrics allows you to check the distribution of various image parameters, helping you quickly filter out problem images and understand the distribution of color, contrast and blurriness in the batch of images.

<figure><img src="https://2101974232-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FO2GXP74UOzykZuHBn8HP%2Fuploads%2FcBbgIXEWIGeH9JtVdH2n%2Fimage.png?alt=media&#x26;token=774a5a96-9c63-4c64-b182-60a840eba909" alt=""><figcaption></figcaption></figure>

<figure><img src="https://2101974232-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FO2GXP74UOzykZuHBn8HP%2Fuploads%2F9VxLFJrYnr6v7vxYMfyI%2Fimage.png?alt=media&#x26;token=7c5234af-0d0a-409c-84bd-8c42fa2df344" alt=""><figcaption></figcaption></figure>

**Data Slice with IQA**

You can also **run IQA (Image Quality Assessment)** directly within any data slice — even if IQA wasn’t executed during the initial import. This makes it easier to recheck image quality, filter blur samples, and prepare clean subsets for training or annotation.

<figure><img src="https://2101974232-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FO2GXP74UOzykZuHBn8HP%2Fuploads%2FhY21lxyhkSY5dcvgi9Ev%2Fimage.png?alt=media&#x26;token=3428f4c8-6963-4951-8a43-40a7bf809183" alt=""><figcaption></figcaption></figure>

<br>

#### Use Cases

* Use IQA to exclude overexposure/underexposure images
* Use IQA to exclude blur images&#x20;

{% content-ref url="../data-management/data-visualization/less-than-use-case-greater-than-clean-raw-data" %}
[less-than-use-case-greater-than-clean-raw-data](https://linkervision.gitbook.io/dataverse/data-management/data-visualization/less-than-use-case-greater-than-clean-raw-data)
{% endcontent-ref %}