Local binary patterns (LBPs) have been fundamental in image processing and machine learning, yet they exhibit limitations such as low preference, generalisation issues, and restricted neighbourhood sampling. To address these, we propose three novel feature extraction techniques: horizontal orthogonal information fused LBP (HOI-LBP), vertical orthogonal information fused LBP (VOI-LBP), and multiplied orthogonal information fused LBP (MOI-LBP). These proposed techniques integrate orthogonal information from the outer 5x5 neighbourhood through logical bitwisefusion with the traditional LBP 3x3 operator. We enhance feature richness and noise robustness by leveraging logical combinations of AND and OR operations on the cyclic 8-bit pairs of corner, horizontal, and vertical pairs of pixels surrounding the central pixel in the LBP’s outer neighbourhood. Experimental evaluations across diverse benchmark datasets, including dog versus cat, CIFAR-10, and MNIST, demonstrate the efficacy of the proposed techniques. We train different deep learning models, such as VGG-16, MobileNet, and EfficientNet B0. The HOI-LBP, VOI-LBP, and MOILBPfeatures consistently do better than regular LBP, showing an average increase of 6.63% training accuracy, and 4.7%in test accuracy. Furthermore, average reductions of 9.46% and 7.51% in training and validation losses signify improvedconvergence and generalisation capabilities. These results show that by capturing subtle variations in texture, gradientenergies, and structural patterns, the fused features offer a more comprehensive representation of image content. Thus,the proposed techniques introduce a paradigm shift in image processing and machine learning, empowering finer detailcapture and improved performance across diverse applications.