top of page

Texture Similarity Analysis Tool

image.png

Tool UI and result visualization

Introduction

Texture Similarity Analysis Tool is a Python-based application designed to automatically detect visually similar textures within large-scale datasets (100+ images).


The tool aims to streamline asset management workflows by reducing manual inspection time and improving consistency in texture organization.

It provides an interactive UI for browsing textures, adjusting similarity thresholds, and visualizing matching results in real time.

​

It provides an interactive UI for browsing textures, adjusting similarity thresholds, and visualizing matching results in real time.

Overview

This project focuses on building an efficient and scalable pipeline for image similarity analysis, combining multiple computer vision techniques with a user-friendly interface.

​

Key Features

  • Batch processing of large texture datasets

  • Single-image query for finding similar textures

  • Adjustable similarity threshold for flexible filtering

  • Real-time visualization of similarity results

  • Modular system supporting multiple algorithms

Tool Design

Tool UI and control pannel

Technical Approach

The system follows a multi-stage pipeline:

  1. Preprocessing

    • Image resizing and grayscale conversion for consistency

  2. Feature Extraction

    • Fast hashing methods: pHash, aHash, dHash

    • Local feature matching: ORB

    • Optional deep features: CNN (ResNet-based)

  3. Similarity Computation

    • Hamming distance for hash comparison

    • Cosine similarity for feature vectors

  4. Result Visualization

    • Display of similar image pairs

    • Interactive UI with similarity scores

image.png

Example of perceptual hash (pHash) compute function​

image.png

File Selection and Input Handling Implementation​

bottom of page