[Paper Review] Content-Based Image Retrieval Using Multiresolution Analysis Of Shape-Based Classified Images

Ibrahim Elhenawy, Kareem Ahmed|arXiv (Cornell University)|Oct 8, 2016

Image Retrieval and Classification Techniques3 citations

TL;DR

This paper proposes a novel content-based image retrieval (CBIR) system that integrates color, texture, and shape features using multiresolution analysis via 2D discrete wavelet transform (DWT) and shape skeletonization. By extracting feature vectors for RGB channels, DWT sub-bands, and skeletal shapes, and storing them in a MySQL database, the system achieves a 96.4% average correct retrieval rate on a 10,000-image dataset with efficient query response times.

ABSTRACT

Content-Based Image Retrieval (CBIR) systems have been widely used for a wide range of applications such as Art collections, Crime prevention and Intellectual property. In this paper, a novel CBIR system, which utilizes visual contents (color, texture and shape) of an image to retrieve images, is proposed. The proposed system builds three feature vectors and stores them into MySQL database. The first feature vector uses descriptive statistics to describe the distribution of data in each channel of RGB channels of the image. The second feature vector describes the texture using eigenvalues of the 39 sub-bands that are generated after applying four levels 2D DWT in each channel (red, green and blue channels) of the image. These wavelets sub-bands perfectly describes the horizontal, vertical and diagonal edges that exist in the multi-resolution analysis of the image. The third feature vector describes the basic shapes that exist in the skeletonization version of the black and white representation of the image. Experimental results on a private MYSQL database that consists of 10000 images, using color, texture, shape and stored relevance feedbacks, showed 96.4% average correct retrieval rate in an efficient recovery time.

Motivation & Objective

• To improve content-based image retrieval accuracy by combining multiple visual features—color, texture, and shape—into a unified system.
• To leverage multiresolution analysis through 2D DWT to extract detailed texture information across multiple sub-bands in RGB channels.
• To classify and represent dominant shapes in images using skeletonization of binary image representations.
• To store extracted features in a relational database (MySQL) for efficient indexing and fast retrieval.
• To evaluate system performance using relevance feedback and real-world image datasets.

Proposed method

• The system extracts color features using descriptive statistics from the red, green, and blue channels of each image.
• It applies four-level 2D Discrete Wavelet Transform (DWT) to each RGB channel, generating 39 sub-bands that capture horizontal, vertical, and diagonal edge information.
• Eigenvalues from the 39 DWT sub-bands are computed to represent texture features in a multiresolution framework.
• The image is converted to binary, skeletonized, and dominant shapes are detected and encoded into a shape-based feature vector.
• Three distinct feature vectors—color, texture, and shape—are combined and stored in a MySQL database for indexing and retrieval.
• Relevance feedback is incorporated to refine retrieval results iteratively.

Experimental results

Research questions

• RQ1 Can a multiresolution texture representation via 2D DWT improve texture discrimination in CBIR systems?
• RQ2 How effectively can shape skeletonization capture and represent dominant shapes for image retrieval?
• RQ3 What is the impact of combining color, texture, and shape features on retrieval accuracy in a large-scale image database?
• RQ4 How does the integration of relevance feedback affect retrieval performance in this feature-based system?
• RQ5 Can a relational database efficiently store and retrieve complex visual feature vectors for real-time CBIR applications?

Key findings

• The proposed CBIR system achieved a 96.4% average correct retrieval rate on a private database of 10,000 images.
• The integration of color, texture, and shape features significantly enhanced retrieval accuracy compared to single-feature approaches.
• Multiresolution DWT analysis effectively captured edge and texture characteristics across multiple scales in each RGB channel.
• Shape representation via skeletonization enabled robust identification of basic geometric forms in images.
• The system demonstrated efficient recovery time, indicating scalability for large image collections.
• Incorporating relevance feedback further improved retrieval precision by refining query results based on user feedback.