semperfidelis0510 / entropy-and-symmetry Goto Github PK

The main objective of this summer project is to investigate the relationship between image textures, entropy measures, and image classification. The project will begin by exploring images with repeating patterns and various symmetries to understand how entropy relates to structured textures. Subsequently, we will apply our knowledge to analyze real-world satellite images of Earth, investigating dependencies on location, time, and zoom levels. Additionally, we will extend our scope to investigate video data, studying the temporal evolution of entropy in moving images. Furthermore, the project will include an analysis of subimages within a larger image to examine patterns of entropy at different scales. This project will provide hands-on experience in Python programming, image processing, and the calculation and application of various entropy measures.

The initial stage will focus on collecting a diverse dataset of images containing repeating patterns with different symmetries. We will use sources like public image repositories or image generation services to obtain images with controlled textures. Emphasis will be placed on selecting images with various symmetries, such as rotational, translational, or reflectional symmetries. Additionally, we will generate images of fractals using known algorithms with controlled parameters. These fractal images will be valuable for later investigations into entropy.

In this stage, our focus will be twofold: developing algorithms to calculate various entropy measures for the images with symmetries and creating any necessary image preprocessing algorithms. We will explore and implement methods to compute entropy of the greyscale representation, entropy of the color histogram, and entropy based on different image transformations. The goal is to have a set of functional and efficient algorithms that can calculate entropy values for the images with symmetries accurately. Additionally, we will design and implement image preprocessing algorithms to ensure that the images are in a suitable format and quality for entropy analysis. Image preprocessing may involve tasks such as resizing, normalization, noise reduction, or contrast enhancement. By the end of this stage, we will have a toolbox of entropy calculation algorithms, along with any necessary image preprocessing algorithms, all ready to be utilized in the subsequent stages of the project.

In Stage 3, we will utilize the entropy calculation algorithms developed in the previous stage to analyze the images with symmetries. Using the calculated entropy values, we will order and classify the images based on their complexity and structure. By assigning semantic meaning to different entropy measures, we aim to understand their significance in describing image complexities. This texture analysis and classification will provide valuable insights into the potential applications of entropy measures in image analysis tasks. We will evaluate the classification accuracy and examine how different entropy measures contribute to distinguishing various symmetries in the images, paving the way for practical use of entropy in image classification scenarios.

In this crucial stage, we will investigate the entropy of subimages within a larger image. We will divide the images into smaller, overlapping, or non-overlapping subimages and compute the entropy for each subimage using different entropy measures. By analyzing patterns of entropy at different scales, we will gain a deeper understanding of how image structure and complexity vary within the larger context. This analysis will provide valuable insights into the role of entropy in capturing localized image features and will be essential for the subsequent analysis.

In this stage, we will expand the project to video data and investigate the temporal evolution of entropy in moving images. We will analyze a collection of video clips and compute various entropy measures over time, exploring how the information content and complexity of images change dynamically. This stage will provide insights into the role of entropy in video analysis and its potential applications.

Having gained proficiency in texture analysis and classification, we will proceed to the final stage of the project. We will apply our knowledge and algorithms to analyze real-world satellite images of Earth. The focus will be on studying the textures and patterns present in the satellite images and calculating their entropy using different entropy measures. Additionally, we will investigate the dependencies of entropy on the geographical location, time of acquisition, and zoom levels of the images. This comprehensive analysis will offer valuable insights into the information content of satellite imagery and its spatial and spectral characteristics.

We will be encouraged to document our progress, insights, and results. Regular meetings will be scheduled to provide guidance, support, and foster collaboration among the students. By investigating repeating patterns, real-world satellite images, temporal video analysis, and subimage entropy patterns using various entropy measures, this project will offer a comprehensive exploration of entropy's role in texture analysis and its application in practical image classification and video analysis scenarios.

To ensure smooth collaboration and organization within the project, we have established some guidelines on how to use this repository effectively.

For adding resources, datasets, or images related to the project, you can directly push the files to the main branch. This allows easy access to shared resources for all project members.

When working on scripts or any code development for the project, we encourage the following approaches:

1. Working on the main Branch: For small changes or quick additions, you can create or modify files directly on the main branch. This is suitable for minor adjustments or small code snippets.

2. Branching Out: If you are working on a more substantial script or feature, it is advisable to create a new branch from the main branch. Name the branch with a descriptive title related to the work you will be doing. This allows you to work on your task independently without affecting the main branch.

For individual walk-throughs or experiments that you wish to perform separately, each student should create their own branch. This helps in isolating individual work and avoids potential conflicts when working in parallel.

When you are satisfied with the changes and additions made in your branch, you can create a pull request to merge it back into the main branch. This way, we can review the changes as a team before incorporating them into the main project.

By adhering to these guidelines, we can ensure a well-organized and collaborative environment for our project. It also helps in tracking and managing the progress of various tasks efficiently.