Recently a huge trend in internet of things and an exponential increase in number of malware are helping malware producers to change malware variants through several automated techniques. Automated techniques may reuse some malware segments to produce variants, and these reuse segments can be helpful to distinguish malware families. Malware variants belonging to same class seem to be much analogous in structure and texture. For this reason, the similarity among malware variants can be used for malware variant family classification. This paper introduces a new malware feature extraction method for capturing local and global properties of images as preliminary features of malware families. The proposed method also reduces the feature dimensions through encoding based feature selection. The experiment is analyzed on three publically available datasets of windows system software. Preliminary experimental results indicate that proposed technique is effective to identify malware family.