Machine-to-machine (M2M) communication is one of the latest technologies to support connectivity among numerous intelligent devices. The intelligence of M2M systems can be enhanced by incorporating visual sensor networks (VSNs) and utilising visual information. The conservation of energy within VSNs is one of the primary concerns for resource constrained scenarios, which can be achieved from targeted threshold based optimisation. However, such optimisation may impact the quality-of-information (QoI), which quantifies the degree to which the visual data is suitable for a given application. To cope with such optimisation challenges, this paper presents a self-reconfiguration scheme for visual sensor nodes to dynamically find optimal configurations as well as guaranteeing satisfactory performance to achieve the given QoI target. The optimisation is achieved by selecting suitable configurations for the removal of feature redundancy which minimises the transmission cost and results in a feasible solution that enhances the energy and bandwidth efficiency for M2M communication. The performance evaluation of the proposed scheme is carried out for different required QoI targets, and it is observed that the proposed scheme outperforms the conventional scheme by providing up to 59.21% energy savings at a QoI target of 30dB.