Template matching is a well-known computer vision algorithm that involves scanning a template across various parts of an image. The template is correlated within this algorithm using a similarity or matching score, such as the Pearson correlation coefficient (PCC). Achieving a more accurate match necessitates searching many regions using the PCC metric, which is hindered by the Von Neumann Bottleneck, resulting in increased energy consumption and delays. Therefore, this paper proposes an energy-efficient, comprehensive memristive in-memory computing architecture for template matching with its physical design, where the PCC computation unit consists of a sensor readout unit, data converters, demultiplexers, in-memory memristive computing array, ADC, running sum module and comparator. The PCC equation is approximated, considering the limitations of the hardware characteristics and application requirements. The proposed approximated memristive in-memory based template-matching scheme demonstrates competitive performance compared to the Von Neumann system and achieves around 678× improvement in the power-delay product. Lastly, a threshold-based optimization strategy is suggested to reduce energy consumption in the application.