Image compression using deterministic Compressive Sensing and Vector Quantization

In Compressive Sensing (CS), the small collections of non-adaptive linear projections of a sparse signal can efficiently help in the reconstruction of the image with the help of the image data sent to the decoder. Although number of CS measurements are small but the result obtained is acceptable because the image data is analyzed by Human Visual System (HVS) which is less sensitive to high frequency information that can be reduced significantly. Compressive Sensing alone cannot achieve much, so Vector Quantization (VQ) has been used with it and its implementation gives efficient results. This paper presents a CS-VQ based image compression algorithm in which a weighted sequence is used to convert the incoming signal into sparse form. The implementation of proposed CS-VQ algorithm gives better results than the existing CS-VQ algorithm in terms of comparison parameters like PSNR and subjective quality of image. The effect of changing the image block size and the contrast of the image, on the PSNR obtained by the proposed CS-VQ algorithm has also been analyzed.