NN filter for Very Low Operational Point (VLOP), targeting complexity of 5kMAC/pixel level, was proposed in JVET-AH0051 and adopted to NNVC. In JVET-AI0107, a new VLOP filter based on LOP3 architecture is tested. In this proposal VLOP filter of EE1-1.5 is modified with reduced complexity input feature extraction. It is proposed to remove application of the DCT transform on sparse input data and further reduce input data volume by aggregating supplementary data into fewer number of planes. As a result, volume of the input data to the NN model can be reduced by around 40% with around 2% kMAC/pixel reduction.