The development of human face image recognition has been fueled by the quick advancement in information technology. Face recognition has recently been effectively used in a number of different fields thanks to computers and information technology. This type of application is crucial to the process of digital forensics investigation because it can identify human face patterns such as eye spacing, nose bridging, lip, ear, and chin contours based on the partial matching of images in 24-bit color image format. The suggested hybrid Lion with Grey Wolf Optimizer (HLGWO) for partial face recognition is presented in this work. The first step is to gather the facial photos from the two public databases FASSEG and ORL. Following that, the Fully Convolutional Network (FCN) and Sparse Representation Classification (SRC) are used to extract the feature vectors from the gathered face photos. This approach is new in that it seeks to optimize the sparse coefficient of Dynamic Feature Matching (DFM), with the goal of minimizing the reconstruction error. Additionally, the structural similarity index metric is presented in this study to determine the similarity scores between a gallery sub-feature map and probing feature map. The dimension of the retrieved features was also reduced or the unnecessary features were rejected using an HLGWO. When compared to the current approaches in an experimental investigation, the suggested methodology performs better.