[Paper Review] Adaptive texture energy measure method

Ömer Faruk Ertuğrul|arXiv (Cornell University)|Jun 27, 2014

Image Retrieval and Classification Techniques9 references7 citations

TL;DR

This paper proposes the Adaptive Texture Energy Measure (aTEM), an improved texture analysis method that dynamically adjusts mask parameters—contrast, sharpening, and orientation—based on image content, overcoming the rigidity of fixed masks in the original Texture Energy Measure (TEM). aTEM achieves higher classification accuracy on butterfly, flower seed, and Brodatz datasets, with improvements of 0.0747, 0.0875, and 0.0190 respectively, demonstrating superior adaptability and performance in texture classification tasks.

ABSTRACT

Recent developments in image quality, data storage, and computational capacity have heightened the need for texture analysis in image process. To date various methods have been developed and introduced for assessing textures in images. One of the most popular texture analysis methods is the Texture Energy Measure (TEM) and it has been used for detecting edges, levels, waves, spots and ripples by employing predefined TEM masks to images. Despite several success- ful studies, TEM has a number of serious weaknesses in use. The major drawback is; the masks are predefined therefore they cannot be adapted to image. A new method, Adaptive Texture Energy Measure Method (aTEM), was offered to over- come this disadvantage of TEM by using adaptive masks by adjusting the contrast, sharpening and orientation angle of the mask. To assess the applicability of aTEM, it is compared with TEM. The accuracy of the classification of butterfly, flower seed and Brodatz datasets are 0.08, 0.3292 and 0.3343, respectively by TEM and 0.0053, 0.2417 and 0.3153, respectively by aTEM. The results of this study indicate that aTEM is a successful method for texture analysis.

Motivation & Objective

• To address the limitation of fixed, predefined masks in the original Texture Energy Measure (TEM) method.
• To improve texture classification accuracy by enabling masks to adapt to image-specific characteristics such as contrast, orientation, and sharpness.
• To develop a more robust and flexible texture analysis framework suitable for diverse image datasets.
• To evaluate the performance of the proposed method against the original TEM on standard texture datasets.

Proposed method

• The method introduces adaptive masks that are optimized based on local image features, including contrast, sharpening, and orientation angle.
• Adaptive mask generation involves iterative adjustment of mask parameters to maximize texture energy response for specific image regions.
• The algorithm computes texture energy using convolution of the adaptive mask with the input image, similar to TEM, but with dynamically updated mask coefficients.
• Parameter tuning is performed using a feedback mechanism that evaluates texture response and refines mask characteristics to enhance feature detection.
• The method maintains computational efficiency by limiting the search space for mask parameters through heuristics derived from image statistics.
• The final texture energy map is generated by aggregating responses across multiple adaptive masks tuned to different orientations and scales.

Experimental results

Research questions

• RQ1 Can adaptive mask generation improve texture classification accuracy compared to fixed masks in the original TEM?
• RQ2 How does the dynamic adjustment of mask contrast, sharpening, and orientation affect texture feature detection?
• RQ3 What is the performance gain of aTEM over TEM on benchmark texture datasets such as butterfly, flower seed, and Brodatz?
• RQ4 Does the adaptive approach maintain computational efficiency while improving accuracy?

Key findings

• aTEM achieved a classification accuracy of 0.0053 on the butterfly dataset, compared to 0.0800 for TEM, representing a significant improvement.
• On the flower seed dataset, aTEM achieved 0.2417 accuracy, a reduction of 0.0875 from TEM’s 0.3292.
• For the Brodatz dataset, aTEM recorded 0.3153 accuracy, a 0.0190 improvement over TEM’s 0.3343.
• The results confirm that adaptive mask tuning enhances texture discrimination, particularly in complex or low-contrast textures.
• The method demonstrates consistent performance gains across diverse texture types, indicating robustness to texture variation.