Evolutionary multi-objective optimization of trace transform for invariant feature extraction

Abstract

Abstract-Trace transform is one representation of images that uses different functionals applied on the image function. When the functional is integral, it becomes identical to the well-known Radon transform, which is a useful tool in computed tomography medical imaging. The key question in Trace transform is to select the best combination of the Trace functionals to produce the optimal triple feature, which is a challenging task. In this paper, we adopt a multi-objective evolutionary algorithm adapted from the elitist nondominated sorting genetic algorithm (NSGA-II), an evolutionary algorithm that has shown to be very efficient for multi-objective optimization, to select the best functionals as well as the optimal number of projections used in Trace transform to achieve invariant image identification. This is achieved by minimizing the within-class variance and maximizing the between-class variance. To enhance the computational efficiency, the Trace parameters are calculated offline and stored, which are then used to calculate the triple features in the evolutionary optimization. The proposed Evolutionary Trace Transform (ETT) is empirically evaluated on various images from fish database. It is shown that the proposed algorithm is very promising in that it is computationally efficient and considerably outperforms existing methods in literatures.

Evolutionary Multiobjective Image Feature Extraction in the Presence of Noise

“…Although there was no significant increase in the identification rate compared to the classical Trace functional [17], the computational cost was reduced because only eight signatures were used instead of 22 in [17]. This paper presents a substantial extension of our preliminary work reported in [24], [25]. New contributions of the paper are as follows.…”

“…Most recently, research effort has been dedicated to selecting the optimal combinations of functionals in Trace transform. In an earlier work [24], [25], we used evolutionary algorithms to find the best combinations of the classical Trace functionals, thereby reducing the computational cost while maintaining high identification performance for RST invariant image identification. Frias-Velazquez et al [26] proposed a feature selection methodology based on Laguerre polynomials by minimizing the dependency among signatures from functionals of the classical Trace transform and tested the method on vehicle identification.…”

A Comparative Study of Multi-objective Evolutionary Trace Transform Methods for Robust Feature Extraction

“…In [8] a reinforcement learning algorithm was applied to the weighted Trace transform (WTT) to find the optimal threshold in the WTT space to minimize the within-class variance only. Recently, in [6], an Evolutionary Trace Transform (ETT) is developed for invariant feature extraction. It has been shown that ETT outperforms the traditional TT in extracting robust triple features from images.…”

“…The main idea is to find optimal combinations of the functionals together with the number of projections in Trace transform to achieve fast and robust feature extraction. It has been shown that evolutionary Trace transform is more robust and efficient than the traditional Trace transform [6]. This paper compares two methods of Evolutionary Trace Transform, method I and II, developed using multi-objective evolutionary optimization of Trace transform to produce candidate features of digital images.…”

“…Evolutionary Trace Transform (ETT) [6] has recently been developed to construct efficient Trace transform triple features to represent an image. The main idea is to find optimal combinations of the functionals together with the number of projections in Trace transform to achieve fast and robust feature extraction.…”