A Review on Feature Combination in Object Classification

Abstract

Research and development could focus on feature combination. It is used in metric learning, text classification, scene classification, disease detection, object classification, and more. Even while SIFT, HOG, and SURF are powerful with some classes, they cannot handle all classes. Inter-class correlation and intra-class diversity make object classification challenging. Using feature combination to combine the strength of numerous complementing features improves multiclass object classification results that are robust to intra-class changes and interclass correlations.

Baseline, boosting, MKL, and other feature combining approaches exist. Even though researching and trying have yielded many important evaluations on these methodologies, more review is needed.