Color Distribution Analysis for Super-Resolution Datasets

This repository provides a set of scripts for analyzing pixel-level color distributions in super-resolution (SR) datasets. The tools are designed to aggregate per-color statistics from image datasets and derive summary metrics such as histograms and average color intensities.

The analysis pipeline consists of two stages:

1. Data aggregation – Extracting and counting pixel color values from images and storing the results in CSV format.
2. Analysis and visualization – Computing statistics and generating visualizations based on the aggregated data.

Before running any scripts, install the required Python dependencies:

pip install opencv-python numpy pandas matplotlib tqdm

To analyze a dataset, you must first generate a CSV file containing pixel color counts. This is done using the aggregate_color_data.py script:

python aggregate_color_data.py <DATASET_DIRECTORY> <OUTPUT_CSV>

• <DATASET_DIRECTORY>: Path to the directory containing dataset images.
• <OUTPUT_CSV>: Path where the generated CSV file will be saved.

Run this command separately for each dataset you wish to analyze.

The analysis scripts assume the presence of the following CSV files, each generated using aggregate_color_data.py from the corresponding dataset:

• q2rtx.csv
• div2k.csv
• realsr.csv

Ensure these files are available in the expected location before running the analysis scripts.

• create_histogram.py Generates color distribution histograms based on the aggregated pixel data.

• calculate_average_intensity.py Computes average color intensity values for each dataset.

• Q2RTX: The Quake II RTX dataset is publicly available on Kaggle: https://www.kaggle.com/datasets/maciejjk208/quake-rtx-v0-3/

Other datasets (e.g., DIV2K, RealSR) should be obtained from their respective official sources and processed in the same manner.