Gray-level invariant Haralick texture features

Abstract

Haralick texture features are common texture descriptors in image analysis. To compute the Haralick features, the image gray-levels are reduced, a process called quantization. The resulting features depend heavily on the quantization step, so Haralick features are not reproducible unless the same quantization is performed. The aim of this work was to develop Haralick features that are invariant to the number of quantization gray-levels. By redefining the gray-level co-occurrence matrix (GLCM) as a discretized probability density function, it becomes asymptotically invariant to the quantization. The invariant and original features were compared using logistic regression classification to separate two classes based on the texture features. Classifiers trained on the invariant features showed higher accuracies, and had similar performance when training and test images had very different quantizations. In conclusion, using the invariant Haralick features, an image pattern will give the same texture feature values independent of image quantization.

Towards guidelines to harmonize textural features in PET: Haralick textural features vary with image noise, but exposure-invariant domains enable comparable PET radiomics

“…When using FBP, exposure must be high enough to avoid negative image values. We also suggest always using greyscale mapping, together with grey level invariant feature descriptions [23]. An additional benefit of adjusting exposure for optimal texture detection will be lower sample size requirement in multicenter studies and, as is becoming ever more important, for machine learning algorithms in medical imaging.…”

“…Several works have examined textural feature stability [20], either under differing image acquisition and reconstruction conditions [21,22], in terms of grey level quantization [23], or under multicenter conditions [24], to identify the least variant features. However, the contribution of image count statistics, a known driver of variability in imaging, remains largely overlooked.…”

Intelligent Estimation of Vitrinite Reflectance of Coal from Photomicrographs Based on Machine Learning

“…The 21 gray-level invariant haralick texture features consist of autocorrelation, cluster prominence, cluster shade, and 18 other features. Detailed information about these features can be found in the work of [36].…”

Optimal Mass Transport for Robust Texture Analysis

“…However, interestingly, it was found in recent studies that this step offers some advantages. It adjusts for heterogeneity in lesion size [14], and produce texture features that are invariant to the quantization gray-levels [15], [16]. We then partition the available data to training and test sets, where the training dataset consists of a subset of MSK slices, and the test dataset is comprised of the whole TCIA slices.…”

“…The method has shown promising results in classifying benign and malignant adrenal lesions. Moreover, considering the GLCM as a density function resulted in texture features that are less sensitive to the gray-level quantization, as shown in [15], [16].…”