There are three main steps to detect fake images, namely feature extraction, feature matching, and feature masking based on the similarity between two or more images, or parts of images. The detection accuracy of fraudulent images highly depends on the feature extraction. Thus, the quality of the extracted features plays a key role in the image forgery detection. Multiple feature extraction methods have been proposed for detecting fake images with different level of success, however, not many have the ability to extract geometrical transformation such as rotation and scaling. Hence, to overcome the issue, this paper presents two separate groups of feature extraction, namely the dimensional reduction and keypoint. Initially, we ran a series of experiments to reveal the best feature extraction methods, involving five methods, from both groups in detecting copy-moved image forgeries. Then in our experiments, we implemented the integration of singular value decomposition (SVD) and speeded up robust features (SURF), discrete cosine transform (DCT) and SURF, and discrete wavelet transform (DWT) and SURF, to formulate a more accurate and robust copy-move detection approach. The best performance of copy-move detection was achieved by deploying DWT and SURF. The integration of DWT and SURF solves the rotation and the scaling issues in copy-move image detection with higher high accuracy and shorter execution time.