Image splicing is the most popular technique of digital image manipulation that blends elements of images to produce deceptive content. The development of advanced detection tools that correlate textural attributes with structural features and color alignment is necessary for research teams to detect counterfeits. Improved feature selection techniques for image splicing forgery detection are produced by combining the Enhanced Flower Pollination Algorithm (EFP) and Adaptive Elephant Herd Optimization (AEHO). Perona-Malik equation processing, which lowers noise and improves picture detail, is the first step in preprocessing. This technique combines ResNet-50 Deep learning analysis with Gray Level Co-occurrence Matrix (GLCM) texture measurements to detect indications of picture spacing fraud. After dimensionality is reduced, the Golden Jackal Optimization (GJO) combination is used to identify the most unique characteristics. The Multi-Class Support Vector Machine (M-SVM) satisfies its function as the comprehensive detection tool by providing the highest classification accuracy. The proposed approach, which combines GLCM texture descriptors, ResNet deep learning features, GJO and JOS feature selection algorithms, and M-SVM classification, surpasses existing benchmarks in accuracy, precision, recall, and F1-score results when evaluated against the CUISDE dataset.