Abstract

Author(s): Suvendu Rup, Nibedita Pati, Bodhisattava Dash

This paper presents an efficient dictionary based image coding technique with static construction of over-complete dictionaries. For training, a static dictionary is adopted which is constructed by a combination of Gabor and Discrete Cosine Transform (DCT) basis function. The training samples are generated from the pixels value of 8X8 blocks of different standard images with varied pixel orientation. Further, the dictionary elements are encoded using orthogonal matching pursuit (OMP) algorithm. The popular transform based image coding is replaced by a dictionary based approach. It has been observed that the dictionary trained with the 8X8 blocks of standard images is capable of extracting the 8X8 blocks from other standard images which are not considered during the training. Coding results are presented to illustrate the performance of the proposed scheme in comparison to best existing coding solutions such as JPEG, JPEG 2000 etc. In general, it is observed that the proposed scheme has superior performance as compared to its competitive schemes